Department of Water Resources Engineering Lund Institute of Technology, Lund University Sweden
Proceedings of the International Seminar Rain water harvesting and management of small reservoirs in arid and semiarid areas an expert meeting within the EU-INCO collaboration HYDROMED (Program for research on hill reservoirs in the semiarid zone of the Mediterranean periphery). Lund University, 29 June - 2 July, 1998
Editor Ronny Berndtsson Sponsoring organizations: ORSTOMIHYDROMED, Swedish International Development Cooperation Agency (SAREC), Swedish Natural Science Research Council (NFR), and Lund University
Report 3222 Lund 1999
Department of Water Resources Engineering Lund Institute of Technology, Lund University, Sweden Coden: LUTVDG/(TVVR-3222)/l-316/1999
Proceedings of the International Seminar Rain water harvesting and management of small reservoirs in arid and semiarid areas an expert meeting within the EU-INCO collaboration HYDROMED (Program for research on hill reservoirs in the semiarid zone of the Mediterranean periphery). Lund University, 29 June - 2 July, 1998
Editor Ronny Berndtsson Sponsoring organizations: ORSTOM/HYDROMED, Swedish International Development Cooperation Agency (SARE C), Swedish Natural Science Research Council (NFR), and Lund University
Report 3222 Lund 1999
Preface The increasing need for water in arid and serniarid countries is putting larger and larger stress on the management system for drinking water, irrigation water, sanitation, etc. In many of these countries, the management systems have traditionally relied on large-scale reservoirs for the collection and storage of fresh-water. Large-scale solutions can, however, not be applied everywhere and for climatic and physiographical conditions. Instead, there is now a tendency to try to find solutions which are based on local pre-requisits and traditional knowledge. In Tunisia, within the framework of the program "Planning of sloping lands, utilization of potential water resources, and maintenance and protection of existing resources" included in the 8th Tunisian governmental plan, the building of 1000 small hill reservoirs by the year 2000 is planned. Out of these, 250 small reservoirs have already been constructed. The objectives are further to reduce soil erosion for farming lands (estimated losses at present 10000 ha/year), to reduce sedimentation in dams (estimated at present 25 M nr'), to increase the groundwater recharge in order to save about 500 M m' of water which at present are lost to evaporation (source: Tunisian Ministry of Agriculture, Direction of Water and Soil Conservation). During 1996 an INCO-DC European Union program was started including Tunisia, Morocco, Syria, Lebanon, France, England, Spain, and Sweden. The objective of this program was to study sustainable management alternatives for small fresh-water reservoirs. The cooperation program, HYDROMED - Research on small reservoirs in the arid zone of the Mediterranean, lead also forward to a specialized seminar on rainwater harvesting and management of small reservoirs in arid and semiarid areas. The objective of the seminar was to exchange ideas and techniques for rainwater harvesting in dry countries as a way to safeguard limited water resources. The final outcome of the seminar is the present proceedings.
3
4
Contents
Page
I. Observation techniques; GIS/remote sensing; climatic, soil, agronomic, and socioeconomic data storage and processing for small watersheds; (chairman: Dr. Jean Albergel, ORSTOM).
9
Hydrology of Sindyaneh Wadi Basin in Syria, Or. Abdallah Droubi, ACSAD, Or. Salah Kara Darnour, MlDS. Or. Jean Albergel, ORSTOM, and Yasser Ibrahim, ACSAD.
11
Small dams' water balance: experimental conditions, data processing and modeling, Dr. Jean'Albergel, ORSTOM, Mr. Slah Nasri, INRGREF, and Or. Mohamed Boufaroua, MAT. / 45 Integrating soil profile and soil hydraulic properties data bases to be used in simulation models and land evaluation expert systems, Prof. Felix Moreno, Dr. D. de la Rosa, and Dr. 1. E. Fernandez, IRNAS.
59
Lebanese hydrology and needs for water storage, Or. Bassam Jaber and Or. Fuad Saad, MHER.
71
Remote sensing applications for the management of small catchments in arid and semiarid areas, Dr. Chuqun Chen, CAS.
85
2. Water quality and quantity; hydrological and transport modeling; (chairman: Dr. Jean Khouri, ACSAD).
93
Water chemistry characteristics in small reservoirs of semiarid Tunisia, Dr. Nathalie Rahaingomanana, ORSTOM.
95
Water chemistry of a small reservoir catchment in central Tunisia, Or. Jean-Pierre Montoroi, Or. O. Grunberger, ORSTOM, and Mr. Slah Nasri, INRGREF.
107
Solute transport and soil water content measurements in arid soils using time domain reflectometry, Dr. Magnus Persson, LU.
123
Decision support system in hydrological modeling, a case study in China, Dr. LinusZhang, LU.
135
5
3. Rainwater harvesting; infiltration techniques and modeling; infiltration and erosion (chairman: Dr. Nejib Rejeb, INRGREF).
149
Water harvesting techniques in the Mediterranean, Prof Dr. Dieter Prinz, KU.
151
The use of TOR for wetness measurements in soil erosion and conservation practices in small watersheds, Mr. Slah Nasri, INRGREF, and Or. Patrick Zante, ORSTOM.
165
Land use transformation impact on reservoir siltation in Morocco: the need for better assessment tools, Or. Abdelaziz Merzouk, IAV.
191
Modeling small dams' siltation with MUSLE, Dr. Jean Albergel and Mr. Yannick Pepin,ORSTOM.
195
Small-scale cistern system for rainwater collection and storage in north-western China, Dr. Linus Zhang, LU and Prof. Kun Zhu, LRI, and Dr. Ronny Berndtsson, LU.
205
Disinfection and fresh-keeping of rainwater in small scale cisterns, Prof. Kun Zhu and Or. Chen Hui, LRI, Or. Linus Zhang and Dr. Ronny Berndtsson, LU.
215
Strategy of soil and water conservation in Tunisia, Or. Habib Farhat and Dr. Mohamed Boufaroua, MAT.
231
4. Reservoir planning, operation and management; Rainfall-inflow relationships; Dam design and operation; Surface-groundwater interactions (chairman: Dr. Abdelaziz Merzouk, IAV).
255
Groundwater recharge and modeling in an experimental catchment, Mr. Slah Nasri, INRGREF.
257
Deterministic versus stochastic hydrological modeling; uncertainties and decisions, Mr. Jan Hoybye, LU.
275
Neural network methodology to simulate discharge, Dr. Cintia Uvo, LU.
295
Appendix I. 2.
Program List of participants
307
311
6
Abbreviations: ACSAD: CAS: IAV: INRGREF: IRNAS:
KU: LRI: LU: MAT: MI-IER: MIDS: ORSTOM: SLU:
Arab Center for the Studies of Arid Zones and Dry Lands, Syria. South China Sea Institute of Oceanography, Chinese Academy of Sciences, China Institute for Agronomy and Veterinary Hassan II, Morocco. National Institute for Research on Rural Engineering, Water, and Forestry, Tunisia. Institute for Natural Resources and Agrobiology, Spain. KarIsruhe University, Germany. Lanzhou Railway Institute, China. Lund University, Sweden. Ministry of Agriculture, Tunisia. Ministry of Hydraulic and Electric Resources, Beirut, Lebanon. Ministry of Irrigation, Damascus, Syria. French Institute for Scientific Research and Cooperative Development, France/Tunisia. Swedish Agricultural University, Uppsala, Sweden.
7
8
Session 1. Observation techniques; GIS/remote sensing; climatic, soil, agronomic, and socioeconomic data storage and processing for small watersheds; (chairman: Dr. Jean Albergel, ORSTOM).
9
10
Rain water harvesting and management of small reservoirs in arid and semiarid areas Lund University, Sweden, 29 June - 2 July, 1998
Hydrology of Sindyaneh Wadi Basin in Syria
Dr. Abdallah Droubi', Dr. Salah Kara Damour', Dr. Jean Albergel', and Mr. Yasser Ibrahim' IACSAD, Arab Center for the Studies of Arid Zones and Dry Lands Division of Water Resources S.P. 2440, Damascus, Syria 2Ministry ofIrrigation Damascus, Syria 3Mission ORSTOM B.P. 434 1004 Tunis, El Menzah, Tunisia
,,,. 12
Hydrology of Sindyaneh Wadi Basin in Syria A. Droubi', S. K. Damour', J. Albergef', and Y. Ibrahim' 1ACSAD, Arab Center for the Studies of Arid Zones and Dry Lands, Division of Water Resources, B.P. 2440, Damascus, Syria. 2Ministry ofIrrigat ion, Damascus, Syria. 3Mission ORSTOM B.P. 434, 1004 Tunis, El Menzah, Tunisia.
1. GENERAL PRESENTATION: 1.1 Physiography: The territory of Syria can be divided into three main physiographic units: 1. The Western Mountain Ranges 2. The Southern Plateaux 3. The Eastern Plains. In the northern part of the country, the western mountains stretch northwards along the Mediterranean coast. The southern part of the western mountain ranges includes the Anti-Lebanon mountains (2600 m) and Jebel Esh-Sheikh (2814 m). The southern Plateaux include the Hauran volcanic plateau in the southwest and the Hamad plateau in the south east. The eastern plains include an arid steppe, and also a semi-arid region with the most fertile land in the country. The steppe includes Badiet-Esh-Sham and Badiet-Er-Rasafa, south of Euphrates river, and Badiet-EI-Jezireh, north of Euphrates river. The semi-arid region includes the Homs- Hama plains, the Idlib-Aleppo plain and the northern Jezireh plains.
1.2 Climate: The climate of the Syrian Arab Republic is of the Mediterranean type, characterized by a cold rainy winter and a dry hot summer with two transitional periods in spring and autumn. The precipitation pattern is influenced mainly by two mountain belts: The western mountain ranges which run northward along coastline and the Tourus mountain ranges which extend along the northern boundary, mainly beyond the limits of the country. The rainy season usually begins in September and ends in April with the possibility of heavy showers in May. High rainfall intensities are recorded in winter in the northern regions, and in spring or autumn in the southern and south-eastern regions. The rainfall distribution in the country is summarized in table (1,2). 13
Table (1)
Annual Rainfall and Evaporation
Name of Basin
Area 2 Km
Average Annual Rainfall Billion mm 3 m 2.658 308
Av. An. Pot. Evaporation mm 1600
1. Barada & Awaj
8630
2. Yarmouk
6724
287
1.930
1700
3. Assi
26446
316
8.357
1400
4. Coastal
5049
967
4.882
1200
5. Tigris & Khabour
21129
402
8.494
1600
6. Euphrates
46416
182
8.448
2000
7. Badia
70786
138
9.768
2000
Average
TOTAL
1900
240
185180
44.537
14
.•3$.E. .........• j
_...
:~~
..
..........-"LE.
..
RE
..
!
i
,oQE
i
.
.~.~.
.Alf····
·'-·1
i
I .1. ...
"':37N
!
!
j.
Average total Precipitation (mm) Yearly In Syria
.....\
Legend ~38N
i····
100 N150
!. I
... ! 35H
..
.. _~
,~fJ
........
f
i i·
..
~
34N
i 33N
i r
L....._ ....
j
It..]
/\/200 /\/250 /\/300
/\/350 '/\'/400 / v /\/500 /\/600
.. \
;,.
Isohyet
!
/\/800
1:\/.1000 1:\/.1200 1\/ 1400
•
City
.. !
.
Figure (1)
··:'001 . . J.~ ~ 32~
N
L\ 15
36.E
...31~
.!SE ...
~E
...E
.4\E...
".E
Average Potential Evaporation (mm) Yearly 137N In Syria
Legend
•
11
C'IJ
'. 'fi
N1200 N1600
I
~
City
Evaporation line
. 38 N
I
. 3SN
IV 1800 2000 'N · '. 2200 rv'2400 le"\',./
,~.........;. •••«•.::.~:.......:•......' l.... ,~
.~
27
3. ENVIRONMENTAL, SOCIAL, AND ECONOMIC IMPACTS OF SMALL DAMS: It is hard to evaluate the environmental and social impacts of the construction of small dams in Syria because of the lack of quantitative environmental and social data prior to dams construction, as well as the difficulties and complications inherent in preparing the studies necessary for that evaluation. The following discussion will merely mention the outstanding effects without using any analysis or statistics.
3.1
Environmental Impacts of Small Dams:
The majority of small dams were built in mountainous areas characterized by very high longitudinal slopes of wadis, or in Badia regions where the surface runoff is characterized by a high suspended and bed load content. After dam construction, it was noticed that the retention lakes have become a sediment trap due to the enlargement of the flow cross section upon entering the lake. This has naturally impeded the sediments from reaching the original areas at the end of the wadi course. The formation of lakes and the subsequent rise up in the water levels in lakes have led to obvious changes in the morphology of the tributaries close to the main course. The erosion and sedimentation regime was substantially changed in the tributaries. The banks of the impoundment lakes were susceptible to scour due to wave formation during wind blowing. This effect could be attenuated by planting suitable plants around the lake or by stone revetment on the weak areas. A lot of dam lakes were planted with fish, but the high turbidity of inflowing water especially in Badia dams has caused fish death. The sedimentation of suspended particles contained in the turbide water has resulted in the suffocation or death of fauna and flora originally existing in the bottom of the dam lakes and led to the formation of bacteria and pollutants. Water stored in the lakes of some small dams is used as water supply for people and for livestock watering. Practice has proven that it is possible to get water of acceptable quality from these lakes jf the following measures are taken: 1. Removal of trees and plants from the lake bottom, and sometimes the removal of the upper soil, prior to storing water and especially before the first filling up. 2. Prevention of any human, industrial, or agricultural waste from reaching the impoundment lake. 3. Control of eutrophication which may be accompanied by disgusting odour, and avoiding or deepening the shallow water areas.
28
4. The use of human, animal, and fish safe pesticides and insecticides to combat disease bearing organisms. Besides the adverse environmental impacts of small dams, there are doubtless positive effects like: 1. Creating a local mild summer climate around the dam sites. 2. Increasing wind blowing over the lake surface and in the area which results in a mild atmosphere. 3. The green landscape (trees. plants, grass,...) in areas that were arid just before building the dams. Such areas were green only for short time after rainfall. 4. Migration of new kinds of birds and animals into the dam area. Such kinds are incomparable with the previously existing kinds prior to dam construction. 5. Reducing air pollution caused by the diesel engines that were used for pumping water from wells. 3.2
Social Impacts of Building Small Dams:
The Syrian climate, characterized by a short rainy winter and a long dry summer, implies providing water in summer either from groundwater or by storing water during winter for using it in summer. It is not always feasible to rely on groundwater which is in many times very deep or of inconvenient quality. Small, dams represent the best alternative which has an advantage even over large dams. Small dams contribute in distributing the national water resources more evenly over the country, while large dams concentrate a huge quantity of water at the dam site and do not permit upstream areas to benefit from that quantity. , The distribution of water resources over numerous areas by small dams had many positive socio-economic impacts, the most important of which are: 1. The provision of domestic water for the inhabitants of Badia and mountainous regions has contributed in improving their hygienic conditions and resulted in a better urbanization level to people living around and downstream of the dam site. 2. The construction of small dams has created temporary jobs during the execution of the dam and its annexed structures, and created permanent work opportunities for the exploitation and maintenance of the dam. 3. The income increase resulting from the irrigation of previously rainfed arable lands has contributed in improving the living and cultural level of local inhabitants.
29
4. Most of the small dams implied the construction of new modern access roads. This has strengthened the mutual relations between urban and rural areas, and led to better level of civilization. 5. Many of the small dams became public tourist centers due to water and green areas. They became also points of contact for farmers from neighboring areas and contributed in developing mutual relations. 6. The amelioration of the conditions of rural areas due to the new water supplies has contributed in the contraction of migration into urban areas. Small dams have also encouraged many Bedouins to settle around their lakes. The adverse social impacts of small dams are relatively few, they can be mentioned as follows: 1. The formation of impoundment lakes has led to the inundation of some arable lands and some houses already built in the area. The land owners were reimbursed, and house tenants were moved into other areas. There were strong objections against dam construction from inhabitants living upstream the dam sites, a complaint atmosphere was created among beneficiaries and damaged people. These problems were settled in the case of dams used for irrigation, the government has deprived the property of both the damaged and beneficiary lands and reallocated the reclaimed land to all land owners. By this way the hurt farmer could obtain a new irrigated land. 2. The new created lakes became attractive swimming centers for local inhabitants. The unsuperintended swimming causes the death of many people every year. In spite of warning, people still swim in the dam lakes; in the meantime, fencing the lakes is quite expensive. 3.3
Economic Impacts of SmallDams:
3.3.1 The money which was spent to transport water from remote water resources into thirsty areas has been saved. This applies for Swaida, Badia, and mountainous areas. 3.3.2 Economic studies have shown that the provision of water supply by small dams is much less expensive than by groundwater wells. The operation and maintenance costs of wells are higher than those of dams. Hard currency that was spent to equip, maintain, and operate wells has been saved. 3.3.3 Small dams have contributed in the attenuation of the overpumping of some irrigation wells. 30
3.3.4 Water stored in some of the small dams is used for supplementary irrigation of cereals in some rain irrigated areas. This has led to more and stable revenue. The adverse economic impacts of small dams are very few, some arable lands were inundated and became no more productive, but this loss is incomparable with the gain of new irrigated lands.
4. HYDROMED PILOT RESERVOIRS: Two hill reservoirs were selected to be as pilot sites for Hydromed project. The two sites are located west of Homs at about 200 Km from Damascus. The two reservoirs are Sindyaneh and Telkalakh hill reservoirs.
4-1 Sindyaneh Reservoir: This reservoir was built in 1967, the catchment area is 4.2 Km2, the dam height is 12 m, and the impoundment volume is 360 000 m3. Several activities have been conducted on this site after it was selected as a pilot site. 1- The spillway threshold was restored and equipped with a staff for reading water levels. 2- The staffs for reading water levels in the lake were rehabilitated. 3- An automatic water level recorder and a rainfall recorder provided by Hydromed were installed. 4- Class A pan fer evaporation measurement was installed. 5- Preparations for measuring sedimentation in the lake were made.
4-2
Telkalakh Reservoir:
This reservoir was built in 1970, the catchment area is 1.75 Km2, the dam height is 10 m, and the impoundment volume is 290 000 m3. The activities that have been taking place after it was selected as a pilot site are: 1- The spillway threshold was restored and equipped with a staff for reading water levels. 2- The staffs for reading water levels in the lake were rehabilitated. 3- The already existing weather station was equipped with class A evaporation pan. This station has the following equipment: Rainfall recorder and rainfall gauge, hygrometer, temperature recorder, and maximum and minimum temperature thermometers.
31
5. PROCESSING OF INITIAL DATA ACQUIRED FROM SINDYANEH RESERVOIR: The Sindyaneh Lake was equipped with an automatic recorder of type PLUVIO-L1MNI 92 recommended by ORSTOM. This equipment can record continuously the variation of the water level in the lake. It is also attached to a rainfall recorder. The data recorded are stored on a memory. Since we have one year of continuous hydrological observation (November 1997-November 1998) a tentative estimation of the hydrological budget of the Lake has been conducted. Using the softwares developed by ORSTOM within the framework of HYDROMED activities which include; HYDROM
: A data bank dealing with data related to water level in the lack, volume stored, surface flooded and outflow on the spillway.
PLUVIOM
: A database dealing with pluviometry, daily and monthly precipitation data.
ARES
: Used to calculate the intensity of rain and capacity of erosion.
SURFER
: Used to calculate curves of thickness/surface/volume based on sedimentation measurements (Bathymetry).
The outputs of this primary study are described below; - Estimation of flood and volume of water stored are shown in table 5. Which was constructed using HYDROM software. On this table we have mentioned only days with precipitation, beginning from 31 December 1997 and up to 30 November 1998. It is clear from this table that reservoir has been completely filled on 28 January 1998, then all the water coming to the lake has not been stored. From this table we can identify the storm event of 6-7 January 1998 where about 102 mm of precipitation was recorded and initiated a flood of 236600 m3. By the end of winter season we had about 434295 m3 of water stored. Figure (6-B) shows an analysis of different rainfall storms reflected as variation of water level in the lake (Hem to the left): we can see that by the end of January, the level of the water in the lake reached the level of the spillway. The reservoir stayed full of water until May, after that the water level and volume in the lake has decreased (Fig.6-A). An analysis of 2 floods arrived on r" January 1998 and 28 to 31 January is also done on figures 7 and 8. 32
Table (5) Analysis of flood in Sindyaneh. Dale Dale
Rain ·fall
Inll.Vol.
mm
m3
Fin. Vol. m3
Spill. flow rn3
Storage
Vol. Flood
Run -off
m3
",3
nun
Qm..
m3/s
Os..,•• l{sl'km'
03/12/1997
60
29260
29970
0
710
710
0.19
0020
5
05/12/1997
0.5 270 21-.5
29260
29730 30450
0
470
012
0.029
8
0
1 430
470 1430
09/12/1997 16112/1997
29020 ,9970
0.38
-0:1"5:; .-----;w
O ~ ------.--s7O -QM - 0023 . ~
31640
..-0:5 0.019 14"0 ~ - 33070 0 -1910 --1910 5 - - - 0 -'.120 --1·.120 ~i -0.004 80 ----nOlO -3-\.,90 I 29-30/1 2/1991 -2"):5 :M490 "-57350 - - C ---2860 '2e60 '075 -'0043 ---li ._~ 0-2850 O:'7S 0027 - ~ GO 37350 40200 3111211997 18/1211997
23/1211997
OS 45200 03/01/1998 04;0111998 ·-1.0 -:-4i4Cu f-Oii:o 71011I9118 . 1ll2.0 ~
10·16/01/1998
41400
0.32
0.019 0.023
6
9.960
2621
-eo
19
361 420
25101/1998
30.5
374900
430630
28/01"998
26.5
430630
.148210
30/01/1998
8.6
-4·\8010 --
'-Y.-:i -M3JOO
06/02/1998
8.0
42r; 490
07/02/1998
51.5
4:>9970
14/02/1998
9
01103/1998
40
17·20/0311998 23/0311998
24103/1098
0
e--o
374900
23-24/0111998
29/0111998
lioo ~
2850 2650 0.75 2eOeSO f - - - O '~ 236600 62."3 570 80570 212 230 '200050 '")G1 4;'0 0 44250
0 0 44
13480
._--
2.140 563 55730 14.7 17580 -'62435 ~ "'39aO -.o.ii 0 -'2-i8JO 65:l . 1.052 ~ii7
---0
--19159
!--:mm - - 0 -3480
3480
0.92
82964
21.8
"'\"3380 19150
22 270;
02·\
0.116
30
1.22
0030
8
4350 8857 233 ---4-0 --2253 '(jS9
0.056
IS
- , 79·10
----ci'l9i --6-'0 "-8550 2Ts ~
'-------00
-
7837'1
434610
434665
841
SS
896
428810
433450
0
4640
4640
6L'.O 428810 17.5 I 434610
,133 160
~
28;03/1996 ---;jij I ~ --m1151J 485 I ~ 434185 29,30/03/1908
4590
-4507 -'
43~6sQ --nTJ .,~
sQ.i 1226 I~ 0.08·' 10.263
434560
-t~5
0.482 ~
J:Ss ----o:IOs ---28
55730
sss
418 OlD ~8Jrj
13480
5
65412
---so -~ ~-
3480
3576
0.94
0.065
735
86148
\7.4
2.499
'------ssB 30·1
17
03·04/04/1998
tS,5
433450
4346·H
17393
119·1
18587
4.89
1.156
13/04/1998
0.5
431 130
432290
0
1160
1160
031
.~
4
21/04/1998
33.5
429970
434540
614
4570
5184
1.J6
0.407
107
6250
1.64
1.583
~
4001
1.05
0059
16
22104/1998
455
434610
435060
5800
450
26-27/04/1998
100
432290
434295
1995
2005
15/11/1998
16.0
26000
26600
0
600
016
-----wli'ii998 ~
25600
26000
0
600 400 --'100
29/1 1/1998
42
25200
25600
0
400
400 ~;
0.056
IS
30/1111998
205
25200
28100
0
2900
0.76
0.222
58
2900
01i
'------ooil ~ -_. -0.038
10
33
Fig. 6-A - Rerelationship between precipitation (in mm) and variation of water level in the lake expressed as H in Cm.
Fig. 6-B - Variation of volume of water stored in the lake from December 97 to 2 Dec. 98.
34
. >".
. '~~rt-~·
SYndu.neb~flow of7 jOn. ~998' "
,
,
::c.-Qe~t,
:-'-IPlt·
>1
,I· .
,.'1:'
. c :"
" ·4
........]: ;': .
.-~,~c~~t~~;;~§~~,~~;~.~r~~l}: .,
.Fig. 7 -Flood
~
'.'
an~'YSis of rainstorm event on7January 1'998..
'~
.
"".'>.. ' ;:"
"
35
~
c
.,1.',
'
'..5·>
,0
.·.· . i.:. ]
S',i: .. .~
.
"
"
...•
'.. -. '.' "'i5~li ""
.':'.\
'20
~'6,,4 ', . . '
~;
'2S'i;~·\t' ,
-c~"" '---;b~.
;oTH", '3S.·
q
v-~/m;·
~:
',. ::. ,
36
These Figures show the relationship between precipitation and quantity of runoff given as Q for a given time. We can see the variation of discharge related to each quantity of precipitation. - A tentative estimation of runoff coefficient has been done in figure 10 where we have as coordinates the precipitation in mm and runoff in mm expressed as thickness of water. The runoff coefficient could be estimated to vary between 30 to 40% , which is in good agreement with values known for such areas.
Conclusion: It is the first time that a study on water budget of small catchment areas could be done in Syria. HYDROMED research programm has facilitated this work by providing the necessary equipment and the scientific support to conduct such study. The processing of initial data acquired from the pilot site of Sindyaneh has shown that the results obtained by using softwares . produced by ORSTOM within the framework of HYDROMED programm, are in good agreement with observation. It was also possible for the first time in Syria to calculate the runoff coefficient for Sindyaneh basin. Monitoring on the site will continue later, and a new campaign to measure the sediments in the lake will be done. Such measurement will help to estimate the rate of erosion on the watershed, which is a major question in Syria.
37
References
HYDROM 3.2 (1994). Guide pratique d'utilisation, Laboratoire d'hydrologie, Montpellier, ORSTOM 88 pp. HYDROM 3.2 (1997) . Gestion et traitement de donnees hydrometriques, Laboratoire d'Hydrologie Eaux continental, ORSTOM. KARA DAMOUR,S. and MISKI,A. (1997). Small dams and hill reservoirs in Syria, ACSAD. 89 pp. PEPIN, Y., LOUATI, M.B. (1997). Rapport de mission en Syrie, Decembre 1997, programme HYDROMED, ORSTOM, 19 pp. PEPIN, Y. LOUATI, M.B. (1998). Rapport de mission en Syrie du 8 au 17 Decembre 1998, programme HYDROMED, ORSTOM, 31 pp. PLUVIOM 2.1 (1994). Logiciel de gestion de donnees pluviometriques manuel d'utilisation, Laboratoire d'hydrologie, ORSTOM. SAFARHY (1994).· Logiciel de calcul statistiques et d'analyse frequentielle adaptes a I'evaluation du risque en hydrology, Manuel de reference, ORSTOM.
38
SINDYANEH .......................... - ....... : 34° 42 =4843 Mollllf:lza' : Hours Latitude
HILL RESERVOIR E Km
Longttnde: 36° 25
N Kill
Y
X
= 377.8
Basin
: Assi
C/um,cteristics ofO,td,,,,ellt Area: 2
:
4.2
KIII
Annual rrccipi.ation:
767
III III
Surface Area
Rainfllll Stations
:
Sadd K. l lunuuum
Years Covered
:
1972-1993
. ..
Aunual Evaporation: Lcngth of Wadi
1370
III III
:
4.2
Kill
Maxfmum Altitnde :
620
III
Minimum AltitlHle :
29
:
III/KIII
Uasalt
Rocks: Soils
m
500
Slopc:
..
Clay
Sediments :
.
Low Suspended Load
.
Soda-Economic Conditions: .. Wdl educated. medium
10
poor people
.
Clltlracteristics ofLuke: Year of Construction :
1967
..
Volnme of Retention (V)
360000
nr'
Surface of Rctcntion (S)
90000
111
2
Ratio (VIS)
4
III
nyke Ileight
12
III
I>ykc Length
Nature of Spillway
512
Lateral
"eight of Spillway
I
Width ul spillwuy
m
12
Maximum discharge of SlIillw:!}'
Diumeter of Outlet Pillc
111
20
III
Irrigalinll
Watcr Quality :
l lydrocarbonatc
J
/Scc nllll
3110 ..
W:,tCI' Use :
N.B. : No Crest tor spillway.
m
.
.
..
39
Charactcrc~lIc curvcs lor dam AI_~undlanch
5U1lace
51U.6 Max Storaqo level __ Storage
suuace (1112.\ 0
20000
40000
OO(jOO ,
511
00000
100000
510
509
508
507·
to
~506
..>
Qi
-1
505
504 .
503
502
501
,
500 0
50000
100000
150000
200000
250000
300000
350000
400000
Storage 1M3)
40
f--
!
I i
I
M~ YE."R: I
! I
I I
7-.... -., ' ... 73_-:-4
7.£ ~s 76 76 77
is
rt
78
78 ig 79 ao
80 8' 81 82 82 83 8384 84 85 85 56 88 87 87 88 8889 8930 90 91 91 92 92 93 9394
Cc;'
DE':
NCV
Fa
JAN
7'3.31 74.41 24.21 85.61 28.4' ss.e. 1~.61 140.5, 'O.::! '~7' Si.'; i~: 1.1, "1; 168.9' 117.5; '56, 56; 102: 1:3.2: , 374 26.21 111.7' 144.41 , 10.6i 78.6/ 1Si9i 29 79.7 153.91 96.81 :51.71 3::.' j 4'. ~ I 19' .91 2e9.9 1 10.::: 1175: 188.5! 116.7' 44.7: 66.6: 49.7: 100.31 3' .81 1':'6' 28.5i 115.1 i 88 -, '00' 49.41 . 85.:, 39,9: 145.: i ' I 15.91 1131 '42.7: 178.5: : :7.61 1261 21.2! 257.91 ; sa.81 '53.91 206.51 68.7: :'loll 11491 '32.2: 61.7: ; 8.61 90.41 161.11 415.51 ,:, .s 88.71 43:.5i 419.91 0; 11731 169.9. S72; l' 1 I 8" i 2961
- ..... (/
..
,.~
,
:
, ,
, ,
I
~.2~
:
I
MAR
APR
130.81
59.71 60.41 55.a! 131.2: :8.5: 28: 30.11 173.61 65.6!
'
';UN
MAY
0.31 0: 9.9 1 27.4; 7.3;
8Ui 63.!1 =.9: 6:3.:: 205.2: 86.31 87." :029: 23.-r 84.8/ 118.91 26.1 : 40.6; 80.81 97.41 170.1 ; 15.:: 1~5, :741 95 i 140.91 73.31 38.:~ 13.5; 38.41 42.' l 50.3! 247.1' 01 77.61 1:9.2! 128.1' 5.4; :C7.' I :g.Z~ 17.7! 9.1 3.2: 88.4' :::'6 ' 01 80.91 ~.81 32: 6.5: 193.5i 200.61 19.5i 63.41 14.5i 10.:: 13.7 1 8.5; 156.1, 0' 38.' , , 7.:~ ~.9; 195.1' 241.5; 0: BC.61 180.91 1:3.3: ""'1:\ .... 14:;:,i: 1G6.41 ~.6' ...... ~l
I
i
!
I
:
I
:
i
:
I
AUG
JULY
:U! 0' O'
0, 7.::! 0; 0, 01
,. ....,. ,
0; 10.~ , 01 0' 4.4; 0
:5.4
1
c· 0:
01
58.i
c
0: 0' O~ 1:'::' 0: 0' 0 0: 0'
0: 0' 0 0 C
0, 01 0: 01 0: 01 0:
I
SE? YE."RL~
0: 0.4:
0 0'
oi
I
01492.1 0.7· 5574 C.9. 5~.:20.1 880.9 0: 769:ii j I
0:
0: s7::":fl
0:
0, 453.7 i Oi1038.41 0: 856.7 I 0: 659.3 ! o 649.3 I ' -._-~ 0,6:0.9 ,
01 0' 0: 0:
0: Q; ;"';1
0· 01
o:
JI
01 0: O!
O'
"-,, ~,~
0, :10 I 01 9598 I 2.:: 880.4 0: 587.7 I O' 545.5J 011420.3 1e.411392.5 01 988
I
I
i
; 757.34
:
,
,
iXo=
:5ecma= C.r-:
::74.~
: 0.::57. _
J
1.1
::
o
lA
J
!
_______ 1\1nYl"llnc~ mc nsuulle s
If)
11)
I I
11
i
IJ
_L __ ~ J F
. 1,\
A
M
I
RE LIA T lYE
BEYROUTH
I
_JI
40
70
(,/;.
01
III
';'1·'
..
60 1;~;~'
~
'.(1
I:K
/
'n
I,ll
';11
i'
\:1
lO
;"1)
10
84 1.·1
A
'-I
n
lA
A
Rain water harvesting and management of small reservoirs in arid and semiarid areas Lund University, Sweden, 29 June - 2 July, 1998
Remote sensing applications for the management of small catchments in arid and semiarid areas
Dr. Chuqun Chen Remote Sensing Group South China Sea Institute of Oceanography Chinese Academy of Sciences 164 West Xingang road Guangzhou 510301, China
85
Remote sensing applications for the management of small watersheds in arid and semiarid areas C.Chen South China Sea Institute a/Oceanography, The Chinese Academy a/Sciences, Guangzhou, 510301, China.
Abstract This paper introduces the general methodology on the remote sensing applications for the environmental management of small watersheds in arid and semiarid areas. The main contents include: a) the available satellite remote sensing data and their characteristics; b) the methods on image interpretation and the classification for thematic mapping; and c) the method of soil erosion modelling for small watersheds, taking the typical small watersheds in the Loess Plateau as examples, from remote sensing data and ground measurement data, and the method of reservoir siltation prediction.
Introduction The satellite remote sensing technology has played an important role in the investigation and management of the environment and the natural resources since early 1970s when the first satellite was launched for the Earth observation. All scientists related to geo-sciences can extract useful information from the remotely sensed data for many purposes, such as, geological mapping, mineral resources surveying, weather report, land use and land cover investigation, land evaluation, urban planning, water quality monitoring, environment assessment and management, crop yield estimation, natural disasters monitoring and their loss estimation, and so on. The remote sensing technology has become a useful and effective tool for geologists, geographers, oceanographers, meteorologists, hydrologists, and ecologists. Comprehensive information is required for watershed management, and the remotely sensed data just right contain the comprehensive information. The main information for watershed management, such as land use, land cover, the topography, soil erosion, the river (water-net) system, gully density, water quality, etc., can be interpreted from the remotely sensed images. The dynamic variation of watershed environment can be monitored by using multi-temporal remotely sensed data. This paper will introduce the properties of related remotely sensed data, the methods on image interpretation, soil erosion modelling and reservoir siltation prediction for management of small watersheds.
The satellite remote sensing data There are plenty of remote sensing data, such as the aerial photographs, the satellite visible and infrared images, and satellite radar images, which can be used for watershed management. The Landsat Thematic Mapper (TM) data and the Spot Earth Observation Satellite SPOT data are widely used for small watershed management.
86
The TM sensors was carried by Landsat 4, launched on 16 July 1982, and Landsat 5, launched on I March 1984, both of which are in Sun-synchronous orbits with equatorial crossing time approximately 9:45 a.m .. The satellite orbits are at an altitude of 705 km and provide a 16day, 233-orbit cycle with a swath overlap that varies from 7 percent at the Equator to nearly 84 percent at 81 degrees north or south latitude. These satellites were designed and operated to collect data over a 185-km swath. The TM sensor has seven spectral bands, which wavelength ranges are from the visible, through the mid-infrared (IR), into the thermal-IR portion of the electromagnetic spectrum (Tab. 1). The TM sensor has a spatial resolution of30 meters for bands I through 5, and band 7, and a spatial resolution of 120 meters for band 6. The TM data provide the surface information of watersheds from the early 1980's to the present, can be employed to monitor the changes occurring on the surface of watersheds.
TM Bands
Table I Bands of Thematic Mapper (TM) and SPOT Data wavelength Wavelength Resolution Resolution SPOT (meters) Bands (urn) (urn) (meters)
I
0.45-0.52
30
I
0.5-0.59
20
2
0.52-0.60
30
2
0.61-0.68
20
3
0.63-0.69
30
3
0.79-0.89
20
4
0.76-0.90
30
Panchromatic
0.51-0.73
10
5
1.55-1.75
30
6
10.40-12.50
120
7
2.08-2.35
30
The SPOT satellite series has been operational for more than ten years. SPOT I was launched on 22 February 1986, and withdrawn from active service on 31 December 1990. SPOT 2 was launched on 22 January 1990 and is still operational. SPOT 3 was launched on 26 September 1993. An incident occurred on SPOT 3 on November 14, 1997. After 3 years in orbit the satellite has stopped functioning. The SPOT 4 was launched on 24 March 1998. The SPOT 5 is scheduled to be launched late in 2002. The SPOT data have 3 multispectral bands, which wavelength ranges are from the visible, near-infrared (IR) portion of the electromagnetic spectrum, and 1 panchromatic band (Table I). The SPOT data have a high spatial resolution, 20 meters for 3 multispectral bands and 10 meters for the panchromatic band. SPOT's oblique viewing capacity makes it possible to produce stereo image pairs by combining two images of the same area acquired on different dates and at different viewing angles. The Stereo image pairs are mainly used for stereo-plotting, topographic mapping, and automatic stereo-correlation, from which Digital Elevation Models (DEM) can be directly derived without the need of maps. SPOT's oblique viewing capacity allows it to image any area within a 900 km swath. Oblique viewing can be used to increase the viewing frequency for a given area during a given cycle. The frequency varies with latitude: at the equator, a given area can be imaged 7 times during the same 26-day orbital cycle. At latitude 45 degree, a given area can be imaged 11 times during the orbital cycle, Le. 157 times yearly and an average of 2.4 days, with an interval ranging from a maximum of 4 days to a minimum of I 87
day. The high frequency viewing is very important for monitoring the variation in some case. On comparison with other satellite data, SPOT data have many advantages: high spatial resolution, high temporal resolution (high frequency viewing), stereo image pairs, which making the data ideal on environmental monitoring and management applications for small watersheds.
Interpretation and classification of remotely sensed image Different ground covers have different characteristics in their size, shape, texture, spectral value and temporal appearance, and the classification from remotely sensed images is based on all these characteristics. For instance, the spectral grey values from water bodies are much lower than that from other kinds of ground covers, that makes it easy to distinguish water bodies from other kinds of ground covers. It is also not difficult to identify ponds from reservoirs or rivers according to their shape, texture and the positional relationship with other land covers. The land covers are different from place to place. A classification system for land covers should in advance be set up on consulting experts. Here is a example of a classification system for land covers (Table 2), and the system can be adjusted from the actual situations of the research watersheds.
Classes
Table 2 Classification system for land covers in small watersheds sub-classes
I crop land
I. I maize; 1.2 cereal; 1.3 bean.
2 forest land
2.1 timber forest; 2.2 mixed forest; 2.3 bush; 2.4 orchard.
3 grass land
3. I sparsely covered; 3.2 dense covered.
4 water body
4.1 pond; 4.2 reservoir; 4.3 river.
5 barren Iand
5.1 rock; 5.2 soil.
6 others
6.1 village; 6.2 road.
The classification can be carried out by interpretation with barren eyes according to the characteristics, such as the colour, shape, texture, etc., or carried out by computers on the base of the grey value or grey combinations of the remotely sensed images. The supervised or nonsupervised classification methods are generally used in computer classification. The accuracy of supervised classification is usually better than that of non-supervised classification for small-scale areas. The classification can provide much thematic information for watershed management.
88
Soil erosion modelling and reservoir siltation prediction Soil erosion modelling
There are several methods to estimate the amounts of soil erosion in a watershed. One method is to draw a soil erosion map which can be interpreted from remotely sensed images on consideration of the erosive factors and their combinations. The topography, the ground slope, the plant coverage and the contents and structure of the soil are generally considered as the main erosive factors. For instance, in the Loess Plateau area within Northern Shaanxi province,three-Jevel classification system was used for soil erosion mapping. The first level was classified, according to erosive forces, into water erosion, gravity erosion and wind erosion. In the second level classification, the water erosion, for example, was classified into 7 grades according to the erosive intensity: feeble erosion «1000 tlkm"a), light erosion (10002500 t/krn':a), middle erosion (2500-5000 t/km"a), intensive erosion (5000-8000 tlkrn"a), extra intensive erosion (8000-15000 t/km"a), strong erosion (15000-25000 t/km':a), and extremely strong erosion (>25000 t/km'·a). Then the total amount of the soil erosion in a watershed can be estimated from the grade map of soil erosion. Another method for estimation of the amount of soil erosion in a watershed is to calculate it from a soil erosion model, which could be developed from remotely sensed data and ground truth data, and can provide more accurate estimation of total amount of soil erosion for a small watershed. Here is a example of a soil erosion model developed for small watersheds in the Loess Plateau of the North Shaanxi province, which is developed from the ground truth data and remotely sensed data of six typical small watersheds. The selection 0/ erosive/actors/or modelling: The erosive factors, including the motive force factors and the affecting factors, which affect the ways of erosive action and the evolution of the motive force factors, are selected for modelling according to the principle of soil erosion dynamics and their function in the process of soil erosion. The runoff-producing rainfall, R, is selected as the motive force factor, and other six factors are selected as the affecting factors. They are the factor of loess contents, S, shown by the coefficient of sand (diameter >0.05mm) to silt (diameter 9:!!JiI -~O O~~~'O
-3,5 -.
o
0
0, ,,0'
gypsum pKs
~
4.58
»:
-4 "
log(SO/) -4,5 . -:._---+---J--f---+--t----I -4,5 -4 -3,5 -3 -2,5 -2 -1,5 -I
Figure 5. Calcite (a) and gypsum (b) equilibrium diagrams for the hillside reservoir water.
Evaporation simulations Although the water chemical evolution is not only determined by evaporation processes, we compared the observed geochemical evolution with evaporation simulations for Fidh Ben Ali, M'Riehet el Anze, and Es Seghir reservoirs which represent the three different geochemical groups. We considered periods where salinity increased. In Fig. 6, we can note differences between the observed and the simulated evolution regarding concentration for Fidh Ben AIL These differences could be expected since, as mentioned before, the actual water evolution may be influenced by secondary inflow, coming from groundwater for example. Moreover, the simulations assume that thermodynamic equilibrium regarding calcite is realised as oversaturation was often observed. In spite of these differences, the measured and simulated E.C. evoJutions are rather close (Fig. 7a). On the other hand, SAR tends to be over-estimated in the model compared to observations (Fig. 7b). This may result from the greater precipitation of calcium in the model, as a result of calcite and gypsum thermodynamic equilibrium. Differences are all the more significant since the water is over-saturated regarding these minerals. The M'Richet el Anze water is quite diluted and only slightly over-saturated regarding calcite, thus, no difference between observations and simulations appears.
103
Fidh Ben Ali observations from /0-94 to 08-95
mM
evaporation simulation
mM
100
(Pco, = /0,3.5 atm)
100
______ 0
x HC03
0---- .........---z
",S04
10
t: s f~ i
~
~"'--- _---~~---Cil=----
+CI
o
oCa
10
oMg oNa
m: x
_
---
»>
0
-~--o~--, 0 ====-;:::;,,~~ Ia----
~G~ .i-> .i->:
G
+
________ +~
FC 1 ' ' - - - - - - - ~-----~----_I x . x---x----~- -·--x'10
FC ~----I
10
0.1 .
0.1
Figure 6. Water composinon evolution with increase of the concentration factor (FC): observations and Expreso evaporation simulation for Fidh Ben Ali hillside reservoir water.
E.C. (dS.m")
a
7 -
SAR 7
6
6
b
0 0
2
Es Seghir (03.96 - 11.96)
M'Richer el Anze (02.96 -11.96)
0
0
I
1.5
2
2.5
2
I FC 0
3.5
I
1.5
2
2.5
FC
3.5
Figure 7. E.C. (a) and SAR (b) evolutions with increase of the concentration factor (FC): observations (empty symbols) and Expreso evaporation simulation (full symbols) for Fidh Ben Ali, Es Seghir, and M'Richet el Anze hillside reservoirs water.
Evaporation simulations allow to determine the theoretical evolution of water quality during dry periods if we assume that no inflow occurs during this period. We carried out evaporation simulations for water with specific initial compositions that allow to delimit the risks related to E.C. and SAR increase, A concentration factor of 5 was considered, it can for example be reached at the end of the summer, in a reservoir having a water level below 2 m at the beginning of the dry season. 104
Results are shown in Table 3 where E.C. values are expressed as quality grades for crops and SAR is presented according to the risk for irrigated soils. As we noted before sulphate dominated runoff presents the greatest salinity and some constraints can exist for irrigation. During the concentration process, the increase in salinity can be limited by gypsum precipitation. Indeed. initial constraints can only be increased and salinity reaching a concentration factor of 5 may be unsuitable for irrigation. Salinity of bicarbonate and chloride dominant runoff is low and entirely satisfactory for irrigation. For a concentration factor of 5, E.C. remains very satisfactory for the bicarbonate dominant water. In the case of the chloride dominant water, E.C. at a concentration factor of 5 indicates that problems may occur for salt sensitive crops. In this group, E.C. is initially higher than for the bicarbonate dominant water and it increases more quickly since Cl and Na do not take part in the salt precipitation. The evolution of the SAR related risk depends on the relative evolution of the cationic proportions and salinity of water. For sulphate water, a high salinity prevents the SAR related risk. For bicarbonate and chloride dominant water, the lowest salinity favours a slight risk. especially in the case of sodium chloride dominant water. Table 3. Hillside-reservoirs water quality evolution during evaporation simulation. E.C. class " SAR related risk Geochemical group FC = 5 FC = I FC = 5 FC = I Sulphate type
A to B
Bicarbonate type
A+
J\+ to A"
without to slight
Chloride type
A+
A' to B'
Slight
* Tunisian
E.C. classes for irrigation:
BloD
Without
n
A ~ all uses « 1.5 g B = not for sensitive crops (1.5 to 3 g 1-') C = tolerant crops only (3 to 5 g 1-') 0= unsuitable for crops (> 5 g 1-')
Conclusion The geochemical characterization of water in the hillside reservoirs of semiarid Tunisia put highlighted different major geochcmical groups with indications about localization and quality of water during the reservoir filling period. The observations made during different hydrological periods helped us to understand the actual geochemical evolution of water and confirmed the importance of reservoir hydrology for this evolution. Expreso simulations were used to detail the risk of water quality deterioration under evaporation for each group. The geochemical and hydrological information acquired for Tunisian hillside reservoirs can now be integrated to bring around concrete actions necessary for the reservoir management. According to above, it would be useful to investigate the nature and volume of secondary inflows. Moreover, exchange phenomena between water and sediments needs to be studied to better understand the geochemical evolution of the stored water. This may be significant for hillside reservoirs where particle transport is important. It may also be interesting to analyse the impact of the reservoirs on the groundwater quality.
105
References Hegelson, H. C., (1969) Thermodynamics of hydrothermal systems at elevated temperatures and pressures. Am. .I. Sci., 267, 724-804. Plummer, L.N., loncs, B.F., and Truesdell, A.H., (1976) WateqF, a Fortran IV version of WATEQ, a computer program for calculating chemical equilibria of natural waters. U.S. Geological Survey Water Resources Investigations paper 76/13. Rahaingomanana, N., (1998) Caracterisation geochimique des lacs collinaires de la Tunisie semiaride et regulation geochimique du phosphore. These de l'Universite Montpellier I, 311 pp. Rieu, M., Vaz, R., Cabrera, F., and Moreno, F., (1998) Modelling the concentration or dilution of saline soil-water systems. European 1. ofSoil Sci., 49, 53-63. Robie, RA, and Waldbaum, OK, (1968) Thermodynamics properties of minerals and related substances at 298.15K (25°C) and one atmosphere (1.033 bars) pressure and at higher temperature. U.S. Geological Survey Bulletin n01259. Rollins, L., (1988) PCwateq: a simple interative PC version of the water chemistry analysis program WateqF, version 2.13. Selmi, S., (1996), Interventions de l'Etat en milieu rural et reactions des collectivites locales face a la gestion d'une ressource rare: Ics lacs collinaires dans le semi-aride tunisicn. These de I'ENSA de Montpellier. Smaoui, A., Camus, H., Albergel, l. et aI., (1996) Annuaire hydrologique des lacs collinaires 1994-1995. Reseau pilote de surveillance hydrologique. Min. de l'Agric., CES/Orstom, Tunis, 140 p. Talineau, l.C., Selmi, S., and Alaya, K., (1994) Lacs collinaires en Tunisie semi-aride. Secheresse, 1994(5): 251-256.
106
Rain water harvesting and management of small reservoirs in arid and semiarid areas Lund University, Sweden, 29 June - 2 July, 1998
Water chemistry of small reservoir catchments in central Tunisia Dr. Jean-Pierre Montoroi', Dr. O. Grunberger', and Mr. Slah Nasri" 1Centre
ORSTOM d'Ile-de-France Laboratoire des Formations Superficielles 32 avenue Henri Varagnat 93143 BONDY Cedex, France 2INRGREF, Route de la Soukra B. P. No. 10 Ariana Tunis, Tunisia
108
Water chemistry of a small reservoir catchment in central Tunisia, preliminary results of water-soil-rock interactions J. P. Montoroil, O. Grunberger ', and S. Nasril lIRD, Laboratoire des Formations Superjicielles, 32 Avenue Henri Varagnat, 93143, Bandy, France. llNRGREF, Rue HMi Karray, HP. 10, 2080 Ariana, Tunis, Tunisia.
Abstract Nwnerous small hill reservoirs have been constructed in Tunisia since the early 1990's. The water chemistry of a representative small reservoir catchment was investigated to elucidate water-soilrock interactions. The groundwater and surface water of the calcareous and marly watershed were characterized by field chemical investigations and pedoJogical observations. The reservoir water was alkaline, with a low concentration, highly oxygenated and weakly carbonated while the ground water was neutral, displayed higher concentration, weakly oxygenated and highly carbonated. Field observations have shown that the reservoir water is infiltrating and supplying a downstream aquifer and that the groundwater is flowing downstream under and inside reservoir sediments. Reservoir water loss was estimated using a conservative tracer and daily water balances. With regard to the chemical composition and the volume of water reservoir, it can be concluded that (i) the reservoir has a minimum infiltration rate of 126 m3 d-I, (ii) above a 4.5 m water level, the daily infiltration rate mostly ranges from 200 to 300 m' d-I, below a 4.5 m level infiltration approximately ranges from 50 to 150 m 3 d-I. Further chemical and isotopic analyses will allow to better understanding of the geochemical processes of the watershed and possibilities to model the water-soil-rock interactions.
Introduction At present, approximately 450 hill reservoirs have been constructed in northern and central Tunisia since the early 1990's. The Direction of Water and Soil Conservation at the Ministry of Agriculture has assigned different aims for these reservoirs: decreasing of soil losses, reducing of dam sedimentation, and replenishing groundwater tables (Albergel and Rejeb, 1997). Water reservoirs should give an opportunity for nomadic families to settle and to find water supplies for agriculture and domestic uses (Talineau et al., 1994; Selmi, 1996). The widespread use ofagricultural practices, such as extraction of groundwater for irrigation and use of fertilizers, leads to a modified groundwater quality. This is particularly the case in Mediterranean climate, where high water demands during the growing season coincides with the dry period (Stigter et aI., 1998).
109
At present, thirty reservoirs are monitored and allow the calculation of water budgets and modelling of catchment water flows (Fig. I). Most reservoirs receive water from surface runoff. The subsurface component of the water balance is usually not a dominant term. From a chemical transport perspective, however, subsurface flows can be important as mechanism of transport for chemicals to and from reservoirs (Winter, 1995). Although hydrochemical investigation combined or not with hydro geological inventory is well established throughout the Mediterranean basin (Armengol et al., 1994; Marc et aI., 1996; Ben Othman et aI., 1997; Petelet et aI., 1997; Stigter et aI., 1998), little is known in Tunisia about the influence of hill reservoirs on groundwater. A chemical and regional typology of hill reservoir water, recently performed by Rahaingomanana (1998), gives a general framework to more detailed studies such as the present one included in the EU sponsored project Hydromed. The main objectives of our work consisted in characterizing spatial water chemistry at a given time (flow and dry periods), in identifying geochemical tracers explaining the relationships between reservoir water and groundwater table and, in fine, in modelling the water-soil-rock interactions. In this paper, we report the preliminary results of a field study carried out during a dry period within and beyond a hill reservoir watershed.
Study site The El Gouazine hill reservoir was chosen among five test sites of the Hydromed project because the water balance is highly negative suggesting an important water loss by infiltration (Fig. I). The catchment is situated in the Ousseltia province, 50 km northwest of Kairouan and more precisely between the Ousseltia and Ksar Lamsa villages (Fig. 2). El Gouazine river is an affluent of Maarouf river and belongs to the endoreic basin ofNebhana river (central Tunisia). The basin outlet forms the Kelbia sebkha located few kilometers north of Kairouan.
o
""" 'J5G .....
Annualralnfall180hyet
o
Monitoredhili resel'Yolr
Maintown 20
Test..lC8(Hydromoo project»
Q IDim!mIiG 2 E15er'I"'lI1 3 fIdtl bell All
11 K,mech 5 .. 'RldJeI .. An:ce
Figure I. Localization of El Gouazine hill reservoir in central Tunisia.
110
The watershed is approximately 18.1 km 2 in area and bordered by SW-NE orientated hills . . Elevations decrease from 575 m above sea level for the highest hills to 375 m above sea level near the reservoir (Fig. 2). With a topographic variation of 200 m and a main river length of 11 km, the average slope is approximately 1.8% or 18 m krrr'. This value is higher than 5% in the upper parts of the valley cross-sections. The El Gouazine region is characterized by a Mediterranean climate with a warm and dry summer, a cool and rainy winter, and highly variable rainfall in autumn and in spring. Mean annual precipitation was 395 mm at Ousseltia during the 1962-1989 period (Bocquet, 1993) and 355.8 mm at El Gouazine during the 1994-1997 period (Guiguen and Ben Younes, 1994; CES/ORSTOM, 1996a; 1996b; 1997). Mean annual air temperature is 19.1QC, with a minimum QC QC of 10.4 in January and a maximum of 28.6 in August (Bocquet, 1993). Potential evapotranspiration strongly exceeds precipitation and is approximately 1680 mm at Kairouan during the 1964-1982 period (Karray and Fakhfakh, 1998) and 1460 mm at Ousseltia during the 1993-1995 period (Riou, 1980; Pernin, 1998). Vegetation originally consisted of Alep pines and Carob trees. Owing to increasing agricultural activities, a large part of the original vegetation has been replaced by rainfed cereals and by irrigated agriculture (mainly tomatoes, peppers, cucumbers, watermelons, almond trees, and olive trees). Alep trees are found only on poor calcareous soils at the top of the hills.
-~R".'
~- ~~:rnll~~
Figure 2. Study site of El Gouazine catchment with geological outcrops. III
Geology and hydrogeology El Gouazine watershed is located on the east edge of a SW-NE orientated syncline, known as Ousseltia syncline and characterized by marly calcareous and gritty deposits (Fournet, 1969). These deposits were raised by tectonic movements in the eastern part of the watershed with a southeastern and nearly vertical dip (Castany, 1951; Jauzein, 1967). The southeastern part of the basin is formed by Eocene sediments which include a mostly carbonated deposit composed of marl and nummulitic limestone from Ypresian or lower Lutetian, and a mostly clayey deposit composed of marl and shelly limestone from upper Lutetian. The northwestern part is formed by Oligocene sediments whose detrital facies is mostly gritty and known as Fortunaformation. Filled up at Quaternary, the syncline deposits have been cut by the rivers (locally known as oued). Gravelly and pebbly colluviums were extensively deposited and ealcreted. Layers of shelly limestone (containing Oyster and/or Gastropod fossils), limy sandstone and marl outcrop in some catchment places (Fig. 2). A regional aquifer is flowing towards the eastern endoreic lowlands (locally known as sebkhas) of the Kairouan plain and is discerned in El Gouazine catchment (Karray and Fakhfakh, 1998).
Soils Most of the watershed soils are highly calcareous and clayey (Brunisso, 1967; Albouy et al., 1995; Bellier et aI., 1997; Montoroi et al., 1997; Bellier et aI., 1998). Calcreted horizons are observed on the summit of most hills. On the hillslope, colluvium is found with a high stone content. According to the World Reference Base for Soil Resources (lSSS Working group RB, 1998), the main soils include calcrete calcisols and calcaric cambisols. Cambisols are mainly formed from marl deposits and locally from limy sandstone deposits.
Dam characteristics The earth and embankment dam was built in 1990. The dyke is 232 m long, 56 m wide, and 10.6 m high, and the spillway overflows at a maximum water level of 8.28 m. The reservoir surface in overflow situation is of9.597 10·' km' for a 18.1 km' watershed surface defining a surface rate of 0.53% and a maximum capacity of233 370 m' (Guiguen and Ben Younes, 1994). In 1997, the reservoir capacity was lower owing to sedimentation, the value being 217 340 m'. The reservoir capacity loss of 16 030 m' in 7 years corresponds to a mean annual loss of2 290 m' and approximately to a maximum 3 m thick sediment deposit (CES/ORSTOM, 1996a; 1996b; 1997).
112
Methods
Waler sampling In May 1998, systematic water sampling was done within the catchment and beyond (Fig. 3). Surface water, groundwater, and reservoir water were sampled and filtered using a 0.2 urn mesh size. Electrical conductivity at 25°C (EC25°c), pH, and dissolved oxygen were measured in field before and after water filtration with WTW devices (LF 330 conductivimeter, pH 340 pHmeter, and OXY 330 oxymeter). Bicarbonate content was assessed in field by complete alcalimetry titration (CAT) with 0.1 N HCI. Filtered water was analyzed at laboratory for major and isotopic elements.
Soil and rock sampling The reservoir was almost empty in May 1998 and bottom sediments were drying. Four pits were dug in the reservoir bottom and two others at the dowstream side of the dam. Soil profiles were described and samples were collected for chemical analysis. Rocks were collected at different places of the catchment close to the wells where groundwater was sampled.
Other data The water level of the reservoir was automatically recorded using an Elsyde device (Guiguen and Ben Younes, 1994). Data were stored in the Hydromed database using Hydrom software (CES/ORSTOM, 1996a; 1996b; 1997). Chemical analyses, carried out by Job et al. (1995) and by Rahaingomanana (1998) for different periods during 1994 and 1995, were used for calculation.
113
catchment scale
Hili reservoir scale
IL,OOlll
0.2
N
Sampling
/
Groundwater ..20 well p4
0
pit
Surface wate, .3 0 fiver + reservoir
0"
•
Pitwithout groundwater \able Waler level gauge
(zeroas reference level) Topographicr8ference level Cl
Emptying water valve
c:J Dam _
Riverbed
N
ReservoirIlmh (overflow level)
Figure 3. Sampling localization of surface waters and ground waters in El Gouazine watershed.
Results
Water chemistry
Surface water and ground water are highly contrasting in the catchment and the reservoir. The reservoir water is alkaline (pH> 10), weakly concentrated (EC < I dS m"), highly oxygenated (> 120 %; range from 9 to 12 mg Lo'), and weakly carbonated (CAT < 2 mmol, L") (Table 1). The chemical characteristics of El Gouazine reservoir water are consistent with those of other reservoirs belonging to the same geological environment. In turn, ground water is close to neutral, more concentrated (2 < EC < 8,5 dS m"), weakly oxygenated « 80 %; range from 3 to 8 mg L"), and highly carbonated (4 < CAT < II mmol, L:') (Table 2). The ground water is less concentrated in the downstream part of the dam suggesting that an upstream ground water flow is diluted by reservoir water.
114
Table 1. Chemical data of surface waters in El Gouazine watershed (data are given for filtered (F) and nonfiltered (NF) samples). Sample
Date
Depth
EC w c
(m)
(ds m")
- 0.1 - 0.1 - 0.1 - 0.1 -0.2 - 0.2 - 0.2 - 0.2
0.859 0.901
pH
T
Dissolved 0,
CAT
~-------
label location
\4/05/98
sI
lake 1 lake 2
14/05/98
s2
marsh
14/05/98
s3
seepage
0~878
0.862 2.845 2.840 7.665 7.700
10.13 9.76 9.93 10.00 6.88 7.23 7.03 7.43
ECwe electrical conductivity measured at 25°C
(QC)
(mg L")
28.7 27.2 29.2 24.3 20.2 25.6 22.5 23.0
12.28
161.8
9.05
122.8
3.85
43.8
7.55
93.9
(%)
(mmolc L~') 2.00 1,22 1.90 1.12 1.72 1.90 10.15 9.20
(NF) (F) (NF) (F) (NF) (F) (NF) (F)
CAT complete alcalimetry titration
Table 2. Chemical data of groundwater in El Gouazine watershed (data are given for nonfiltcrcd samples). Date
Well (w) or pit (p)
Depthl I) EC"'e
pH
T
Dissolved 0,
CAT
Deplh(2)
-------
label location
(m)
(dS m")
(QC)
(mg L-l )
(%)
(mmole L " )
(m) 3.30
._-------------~------
Dam downstream 17/05/98 wl River bed
2.220 2.450 2.130 2.130 2.990 2.990 1.888 2.720
7.06 7.15 7.04 7.08 7.10 7.11 7.38 7.05
19.5 18.\ 17.8 17.3 17.7 17.7 18.7 17.2
4.40 7.15 2.94 2.54 4.57 4.18 3.87 3.73
56.4 80.3 33.3 27.4 42.9 45.5 43.4 41.2
5.20 6.20 6.30 6.60 6.80 7.10 4.00 6.20
8.300
7.02
17.0
6.45
66.5
10.65
2.70
2.240
6.98
20.0
3.80
42.5
6.15
1.45
1.212
7.34
16.1
2.60
26.7
4.50
8.40
3.240 3.250 0.753 3.230 3.230 2.020 4.910
7.13 7.10 7.61 7.26 7.30 7.72 6.85
19.3 18.1 21.7 17.1 17.2 25.6 16.8
7.48 6.92 8.75 5.56 7.54 5.28 1.26
84.8 78.1 108.0 62.1 83.2 70.2 14.8
5.30 5.00 3.70 6.90 6.00 5.20 8.40
3.15
4.180 0.833 0.835 QC EC".e electrical conductivity measured at 25 (I) sampling depth below groundwater level
7.23 7.26 7.25
18.4 17.0 16.8
5.51 3.30 3.74
65.7 37.2 39.8
5.60 6.50 5.70
21/05/98
w2
River bed
21/05/98
w3
River bed
21/05/98 21/05/98
pi p2
Footdyke River bed
Witllin the watershed 16/05/98 w4 River bed 16/05/98
w5
16/05/98
w6
Calcareous outcrop El Aafou
23/05/98
w7
Souk
16/05/98 21/05/98
w8 w9
Fountain Head
23/05/98 22/05/98
p4 p6
Sediment Sediment
Beyond the watershed 17/05/98 w 10 Bou Haleb 22/05/98 wll Larbi
- 1.0 - 15.0 - 1.0 - 2.5 - 1.0 -2.3
- 1.0 - 4.0 - 1.0 - 4.5 - 1.0 - 1.6 - 1.0 - 5.5 - 1.0 - 1.0 - 8.3
- 1.0 - 1.0 - 4.0
2.70 1.70 2.50 1.85
2.00 2.30 3.35 4.00
CA T complete alcalimetry titration (2) groundwater level referencing to soil surface
liS
Pedologicalobservations Reservoir sediments are composed of fine clayey layers alternating with coarse sandy layers. They are overlying the old river bed deposits composed mainly of coarse calcareous gravel and pebble. The sediment thickness is approximately 3 m at the p4 pit confirming the measurements carried out by sounding the reservoir water (CES/ORSTOM, 1996a; 1996b; 1997). Within the gravelly and pebbly layer, a ground water table is observed at about 2.5 m depth (Fig. 4). As shown at the two upstream reservoir cross-sections, a sandy layer is situated at the left side embankment of the reservoir and was overlying the gravelly and pebbly layer. Concerning the reservoir cross-section situated more upstream, the sandy layer is nearly 1.5 m thick, the base being at a 5 m level. In turn, concerning the wider reservoir cross-section, a sandy layer of I m thickness is present in the sediments. According to these facts, we can assume that the high permeability of the sandy layer explains partly the high water loss of the reservoir, especially when the water level amounts to more than 4 m. - _ Cross-section 100m PlC • -:;:::} Overflow IImiC River
Ll00m p2
N
/ _
Sandy layer
It"i:Pt, Reservoirsediment
MS Colluvlum
tH:l Gravelly layer ~ Calcareous mar. p3
I
Pit
Overflow - - - level Groundwater •• - _. table
Figure 4. Deposit structure of El Gouazine hill reservoir (May 1998). The measurements of water level, carried out in the pits and the wells, define a hydraulic gradient from upstream to downstream (Fig. 5). It seems relevant to conclude that an upstream ground water table is flowing downstream under and inside the sediments confirming the assumption suggested by the chemical data.
116
Upstream
Dam
Downstream
10
p'
I
Pit
Figure 5. Groundwater level in upstream and downstream parts of El Gouazine dam (May 1998).
Infiltration rate estimation ofreservoir water by chemical tracer method During a dry period, we considered the temporal evolution of the reservoir volume (V) and the concentration ofa conservative element (Cl"). We compared the reducing volume factor Fv and the concentration factor Fe calculated from an initial given time (i) to a final given time (f) by the relationships Fv = V, / Vr and Fe = CI",/ Cl-, Ifboth factors are equal, there is no infiltration or the ground water inflow compensates the ground water outflow. In turn, the reservoir loses water ifFc>Fv (outflow) and gains water ifFc'''':··1
~
sPreading
measures Mgoud
Well
Vnde,.-qrnund c i s t e r n
8.28 m). After construction of the contour ridges, the lake was dry in July 1997 and midApril 1998, and the water level in the reservoir was not higher than 5.5 m, with 3 m of sediment. Only the 79.4 mm rainfall on September 24, 1998 brought the level up to 7 m, representing a height of 4 m of water in the reservoir.
261
10
140000
126000 1120007
.
96000
j
1
84000 ; j---j--I-J 70000 I~'+------i-+l 56000
I~""'-'I--l 1I
·93
mai- JUII93 93
oct -93
28000
14000 i-L 0
>7-_.-
-r----t =f=H=f=1~
I I I III
.... 11c:
'" 11'" 10
e Q.
c:
0
~t.-~~Ij I I-H-h
.
I
I
'1
I
[
-
'-S
-tit
-r-- li=ll
0
(J)
0
Figure 4. Solution process variance for three different input correlation scales. The input variance q, = I, which implies that the curves represent the error amplification ratio. If the model coefficient, however, is random, either a random process or a random variable, then the solution process variance naturally is increased, see Fig. 5. The two curves ('random variable' and 'stochastic process') shown in the figure describe the effect of added variance due to the uncertain parameter compared to the deterministic parameter white noise curve which is the same as in Figs. 3 and 4. The importance of the choice of how to model the system parameter /lis only significant for parameter values less than 0.5. In other cases, the effect is only of minor importance. In general, the variance equation for the solution process of a linear stochastic differential equation can be expressed as: (2)
289
where it is assumed that the input process Y is a white-noise process (q y = 2(J"~ Ay)' The function g(X) depends on how the parameter fJis modelled and is shown in Table 5. If fJcan be considered
to be deterministic, the parameter variance contribution, naturally, is zero and only the uncertain input process influences the output variance. Table 5. Effect of added variance due to parameter uncertainty in first-order differential equations. The input process is taken to be a white-noise process.
Equation parameter model Deterministic Random variable
Effective model parameter
Random process
fJ-qp/2
Function g(X)
fJ fJ- qp/fJ = fJ(l- CV;)
The effective model parameter in Table 5 reflects the inherent bias in the determination of random parameters in models based on ordinary differential equations. In other words, the randomness in the estimate of the parameter is manifest by a reduction below its mean value. Consequently, the random differential equation can be regarded as an equation with only random input if the effective parameter is substituted for the originally random parameter. - - - - - - _ .. Model parameter fJ 0.Q1
0.1
l'l C
'" > '"
'C
~
Q.
----~~:+ . ---+-+-+
C
o
S (; en
1 -.......
_ _ _ _ _ --'- _ _-'----_--"_...L.._.L._, _
_ _ _ _ . ....L._----'_~_~~~____'__...L
0.1
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _....J
Figure 5. Steady-state solution process variance for the random parameter case. Data used in the graph is: X" = YlfJwhere Y = 0.25, Var{fJ} = 0.09, and Var{ Y} = 1. The curve for the deterministic parameter situation corresponds to the white-noise curve in Fig. 3. The error propagation is illustrated by the following analysis of the steady-state component of a stochastic differential equation. Based on the results in Table 5, the solution process variance 290
for the steady-state case can be normalised by dividing the variance by the steady-state level X" hence obtaining the coefficient of variation
2 CV 2 ==!!..!x X ss2
CV 2 2 fJ4'Wss
CV"
P
2/P
2
== y2 __Y_2 + _fJ_ == _(CV 2 + CV2)
1---
. . 25r-l~j_-a~ --c=05 -c=1 1.5 ------c=2 ..-.- - •• c - 0
~
j------- ---
20 - _
J!l
._
e o
.~
Cl
'" ~
~
~
(3)
p
y
J
~
--
1 _ _-
-
.- -' .. -' ..-=--::-.;. -. _
__
..
I" ,.~::: ,~: : :- -;;f~-~--~;'r'"·==--f :::: ~> =t ~~ j --1- J 1.0
l
j
00
-
o
-- 02 04
0.6
.. - - - - - - - - - - - -
0.8
1
1.2
14
1.6
1. 8
2
Model parameter p
----_.
Figure 6. Error propagation factor r/J for the first-order stochastic differential equation. The parameter fJ is in the above case modelled as a random process. Dividing by the squared coefficient of variation of the input process and taking the square root, the error propagation factor r/J is defined as (4) where c is the ratio between parameter and input CV's squared. This important, yet very simple result, relating relative input and output uncertainties oflinear systems, is shown graphically in Fig. 6 for different parameter-input process uncertainty ratio's. This simple relationship is fundamental in linear random systems, and can be used as an indicator of how much input errors will be amplified or reduced in a general linear system. 291
General conclusion At first sight, stochastic modelling and especially stochastic differential equations describing dynamic systems present a complex problem, and it is true that the development of solution processes is intricate. The solutions to simplified but relevant situations and assumptions, however, are on surprisingly simple forms. Thus the principal conclusion of the paper is that the stochastic approach in hydrological modelling adds complexity to the analysis and application of model simulations, but simple and useful results can be obtained to estimate the variance of functional relationships and stochastic differential equations. Even though the problems presented in this paper can be complex, there appears to be three areas of potential benefits. These are the following:
Decision problems. As was explained earlier, current procedures that use the model with expected values do not yield the correct output values on the basis of the expectation of the model. Furthermore, since neither input data nor model parameters are deterministic properties, errors will be a propagated into the model output. The bias and inaccuracy implied by uncertain input and parameters will affect the decisions and account for misinterpretations. Building simulation models. As hydrological models become larger and more complex, the information surrounding parameter values often decreases. The increased uncertainty in the model output should be analysed. Conceptually, there is a trade-off between a complex model with many hard to measure parameters and simple hydrologic models. Uncertain data in tum complicates the identification or refinement of model processes, and introduces a risk for the inclusion of minor processes or exclusion of important processes. Data collection and value of information. An analysis of the type to be proposed here can indicate how input uncertainty and different parameter values affects the uncertainty in model outputs. Consequently, when collecting data, more effort can be put into those input variables and parameters that reduce uncertainty in the model output the most.
References Burges, S.J. and Lettenmaier, D.P. (\ 975). Probabilistic methods in stream quality management. Water Resour. Bull., 17, 115-130 Beven, K. (1993). Prophecy, reality and uncertainty in distributed hydrological modelling. Advances in Water Resources, Vol. 16, pp. 41-51. Ditlevsen, O. (1981). Uncertainty Modelling. McGraw-Hill Book Co. Environmental Resources Ltd. (\ 984). Prediction in environmental impact assessment. E.I.A. Publication series no. 17. The Hague: Ministries of Housing, Physical Planning and Environment and of Agriculture and Fisheries. Environmental Resources Ltd. (1985). Handling uncertainty in prediction. E.I.A. Publication series no. 18. The Hague: Ministries of Housing, Physical Planning and Environment and of Agriculture and Fisheries.
292
Freeze, R.A. (1980). A Stochastic-conceptual Analysis of Rainfall-Runoff Processes on a Hillslope. Water Resources Research 16(2),391-408. Heybye, J. (1998). Uncertainty Modelling in Hydrological Applications. Ph.D. thesis, report no. 1024, Lund University, Sweden. Lund University (1996). Flood Analysis and Protection System (FLAPS) - Application to the Hongru River. Report 3187, Dep!. of Water Resources Engineering, Lund University, Sweden. Danish Environmental Protection Agency (1978). Simple Lake Water Quality Models (Simple Semodeller). Miljeprojekt No. 16, Danish Ministry of Environment (In Danish). O'Connell, P.E. (1986). Uncertainty if the modelling of hydrological systems. Inst. Phys. Conf. Ser. No. 80, lOP Publishing, Bristol and Boston Sas, H. (1989). Lake restoration by reduction of nutrient loading. Academia Verlag, Richarz Smith, L. and Freeze, R.A. (I 979a). Stochastic Analysis of Steady State Flow in a Bounded Domain, I. One-dimensional simulations. Water Resources Research 15(3), 521-528. Smith, L. and Freeze, R.A. (1979b). Stochastic Analysis of Steady State Flow in a Bounded Domain, 2. Two-dimensional simulations. Water Resources Research 15(6), 1543-1559. Yen, B.C. (1986). Stochastic and Risk Analysis in Hydraulic Engineering. Water Resources Publications, Littleton, Colorado
293
294
Rain water harvesting and management of small reservoirs in arid and semiarid areas Lund University, Sweden, 29 June - 2 July, 1998
Neural network methodology to simulate discharge
Dr. Cintia Uvo Dep. of Water Resources Engineering Lund University, Box 118 S-221 00 Lund, Sweden
295
296
Neural network methodology to simulate discharge C. B. Uvo Department ofWater Resources Engineering, Lund University, S-221 00 Lund, Sweden.
Abstract The observed annual variability in precipitation and water availability in the Amazonia, located in northeasternSouth America, has been shown to be influencedsea surface temperature (SST). However, the links between the large-scale SST patterns and local and regional runoff patterns are essentially complex and still not fully understood, The processes involved are believed to be highly nonlinear, spatially and temporally variable and not easily described by simple deterministic models. Artificial Neural Networks were used to develop models to forecast discharge one or two seasons in advance at 10 sites in Northeastern South America from Pacific and Atlantic sea surface temperature (SST) anomalies. Results were very encouraging.The correlationcoefficient between observed andestimated discharges reached values as high as 0.96.
Introduction During recent decades the study of coupling processes between atmospheric and hydrological scales has become increasingly important (e.g., Eagleson, 1986). This has lead to great advances in the understanding and modeling of regional and global scale hydrology [e.g., Brubaker et al., 1993; 1994). The links between large-scale atmospheric motion and local and regional runoff patterns are, however, extremely complex and still not fully understood. The processes involved are believed to be highly nonlinear, spatially and temporally variable and not easily described by simple deterministic models (e.g., Hsu et al., 1995). The Amazon Basin is the largest river on Earth and it holds about one sixth of the global river water (Dickinson, 1987). About half of the 2-3 m annual rainfall is recycled by evapotranspiration while the rest discharges into the Atlantic Ocean (Molion, 1975; Salati and Marques, 1984; Junk and Furch, 1985; Salati, 1986). Interannual rainfall and discharge in the Basin have been shown to vary considerably from year to year under the influence of sea surface temperature (SST) over the Pacific and Atlantic oceans and climatic phenomena such as El Nifio-Southern Oscillation (ENSO). Rao and Hada (1990), Ropelewski and Halpert (1987), Ropelewski and Halpert (1989), Marengo (1992) and Marengo et al. (1993) examined the response of the region's precipitation to El Nifio-Southern Oscillation (ENSO) events and pointed out that warm/cold El Niiio-Southern Oscillation events are related to below/above nonmal precipitation. Authors such as Moura and Shukla (1981), Hastenrath and Greischar (1993) and Uvo et al. (1997), showed by observational and/or modeling studies that the Atlantic anomalous SST meridional gradient play an important role on the precipitation falling in northeastern South America. Marengo (1992) pointed out that high Rio Negro levels are related to anomalously cold surface waters in the tropical North Atlantic and high SST anomalies south of the equator. Additionally, he showed that abundant rain during the wet season 297
in northern Amazonia is associated with cold surface waters in most of the tropical North Atlantic and eastern Pacific. Molion and Moraes (1987) studied the correlation between the Southern Oscillation Index (SOl) and the discharge for rivers in different parts of the Amazon Basin and found that the SOl and discharge show a good correlation in the eastern part of the basin given a Jag of three months. In spite of comprehensive researchefforts mentionedabove, quantitativemodels to forecast long-term discharge in the Amazon Basin are still lacking. A main reason for this is that the complex physical connections between the general regional climate and the basin's local hydrology are still not well understood. Uvo and Graham (1998)developeda linearstatistical modelto forecast dischargefrom SST for differentsites in the Amazon,Tocantinsand Orinoco Basins.The resultsobtainedby that work lead to the conclusion that it should be possible to forecastdischarge for those sites from SST, but further efforts should be made to improve the results. Some extensivelong-termclimaticand hydrological databases exist,however,for the area(e.g.,Uvo and Graham, 1998). This work presentsthe development of nonlinearmodels using artificial neuralnetworks (ANN) to forecast discharge at some site in Northeastern South America from Pacific and Atlantic oceans SST.
Methodology
The ANN is a nonlinear mathematicalstructure which is capable of representingarbitrarily complex nonlinearprocesses that relatethe inputsand outputsof any system(Hsu et al., 1995). The ANN provides bettersolutions than traditional statistical methodswhen appliedto poorly definedand poorlyunderstood complex systems that involve pattern recognition(poff et al., 1996). It is a viable technique to develop input-outputsimulations and forecastmodels for situations when the objective is an aecurate forecast (Hsu et aI., 1995). Hill et al. (1996) compared neural networkmodels with traditional statisticalmodels for different time series,and concludedthat neuralnetworkmodel forecasts are moreaccuratewhen the data sets have less than 50 historicaldata points, which is the case of the data availablefor this work. Steps necessary to develop and train an ANN involve: a) choosing a pair of data sets that are representative of the phenomenonto be describedand forecasted; b) defininga suitablenetwork(number of layers and number of neurons in each layer); c) training the network to relate the inputs to the corresponding outputs estimatingthe ANN'sweights;and d) validating the identified ANN. If compared to a conceptual model, step (b) is equivalent to the development of the model and step (c) is the calibrationof the parametersof the designed model.
298
In this study, a two-layerANN was used with TANSIG neurons in the hidden(recurrent) layer and PURELIN neuronsin the outputlayer. The TANSIGis a hyperbolic tangent sigmoid fimction givenby: a==(en - e")/(en + e") and PURELIN is a linear function given by: a = n (Hagan et al., 1995). This configuration allows the network to approximate any function with a finitenumberof discontinuities (Dcmuthand Beale, 1994). TIlechoiceof the numberof neurons in the hiddenlayerwasmade through test runsthatstartedfroma smallnumberof neuronsand gradually increased the network size. An optimal relationship betweenthe accuracy achieved and the timespentduringthe training wasfound with40 neurons in the hiddenlayer. We usedback-propagation as training method. 'This is an iterative technique thatinvolves computing the error between the known desired output and the computed neural network output. Based on the magnitude of the error,it adjuststheinterconnection weights in a backward sweepthroughthe network (Frenchet al., 1992). As it employs a gradientsearchstrategy, its performance is quitesensitiveto the startingweights (Hsu et al., 1995). To trainthe network, the routine "trainelm" available in the Matlab Neural Network Toolboxwasused. Training continued until either the sum squared error(SSE)goalwas obtained or a maximum number of iterations was completed. In general, about 150 iterations were enoughto reachthe chosen SSEof 0.01. The trained ANN shouldbe ableto estimate the March-April-May (MAM) average discharge at 10 sites in Northeastern SouthAmerica from seasonal averaged SSTanomalies. Trainings weremadeseparately for each riverstationproviding to the ANN different training setscomposedby a pairof data sets that includedthe dischargeseries (target valuesor output)and different SST periods (input) described as follows: Seasonal averageSST anomalies forthe four seasonspriorto the MAM dischargeseason,i.e., March-May, June-August, September-November and December-February, hereafter referred to as MAM-JJA-SON-DJF. In this case,the information aboutthe SST duringthe entireyearprior to the maindischargeseasonis provided to the network. b) Seasonal averageSST anomalies forthe three seasons notimmediately beforethe MAMdischarge season, i.e.,March-May, June-August, September-November, hereafter referred to as MAM-JJASON. In thiscasetheestimation of the discharge couldbe made3 monthbefore thestartof the main runoffseason. c) Seasonal averageSST anomalies for December-February (DJF). d) Seasonal averageSST anomalies for September-November (SON). The validation of the training was made using the "Ieave-one-out" cross-validation technique. This techniqueconsistsof removingonerow at a time fromthe inputpairof data sets,re-training the ANN and estimating the missingdata. Thecross-validation followed the methodology proposed by Derkset at. (1996).They base their approach on the assumption that local error minimumcan be avoided by choosingvariousrandomstartingweights for the training. In our case, the cross-validation was made using 10 different re-initializations for the training at each removed pair from the input.
299
The correlation coefficient between normalized observed discharge and normalized discharge estimated by the cross-validation and the sum squarederror (SSE)of the estimatedserieswhen comparedto the observedserieswere used to analyzethe resultsfromthe cross-validation. As pointedout above, the cross-validation used repetitions of the same trainingfollowing Derks et al. (1996), starting from differentstartingpositionson the error hyperplane chosen randomly. To better understand this process, we anaIyzed the distribution of the outputsgenerated fromthe differentstarting weights for a same trainingset. One row of the trainingset was removed and the ANN was trained50 timesusing this new input data, startingfromdifferentstartingweights. The removedrow was estimatedfor each of the 50 times. The difference betweenobservedand estimated discharge was calculated foreach of the 50 estimations and the distribution of these differences was analyzed.
Data
The pairof datasets used to trainthe ANN (training set)contained seasonal averaged SSTanomalies and seasonal averagedischarge. The SST anomalies werechosento represent elimatevariability inherentin phenomena such as ENSO and the meridional SST gradientin the Atlantic Ocean.The SST anomalies were used as inputand the discharges as output.
SST
The SST anomalieswere extractedfrom the UWMlCOADS SST anomalies(da Silvaet al., 1994)for the period 1945 to 1991 on a 5° longitude by 2° latitude grid from 300S to 30"N for the Pacific and Atlantic Oceans. Two subsets of thesedatawereutilized. Theyconsisted of the Equatorial Pacificsubset Cl OOS to IO"N) and the Tropical Atlantic subset (300S to 300N). In both cases, the SST data were seasonally averagedfor March-May, June-August, September-November and December-February.
Discharge
The dischargedata consistedof monthlyaverageanomaliesfrom 1946 to 1992 for six stations in the Amazon Basin,one stationin the Araguari Basin,two in the TocantinsBasin and one in the Pamafba Basin(Figure 1).These time serieswereprovidedby ELETROBRAs (CentraisEletricasBrasileiras), ELETRONORTE(CentraisEletricas do Norte do BrasilSA) and DNAEE(Departa-mento Nacional de Aguas e Energia Eletrica). All data series were homogenized by the respective institutions using different methodologies dependingon the data source. Averageseasonal dischargesfor MAM were calculatedand used as targetvaluesin the trainingset of the ANN.
300
i l-,.... . 10'N I
.
o
RI••r
[
h-i""l
':'"'.
I I
S~t10n
RI..r BasinBoundary
~. "'. "'.'
"","'-
0-'I
"g
!
10'Sl
I 70'W
6O'W
SO'W
40'W
Figure I. Map showingthe Amazon,Tocantinsand Parnafba basin and sub-basins. The basins of the Negro, Araguari,Uatuma, Curua-Una, Xingu, Jamari,ParnaIba, Trombetasand TocantinsRivers are delineated by thinnerlines,countryborders are indicated by dashed lines.Numbered circles indicate river dischargestations: 1-Balbina/Uaturna River;2- Belo Monte/Xingu River;3- Boa Esperanea/Pamaiba River;4- CoaracyNuneslAraguari River;5- Curua-Una/Curua-Una River;6- ManauslNegro River;7Porteiraffrombetas River; 8- SamuellJamari River; 9- Serra Quebrada; 10- Tucurui (9 and 10 at TocantinsRiver).Modifiedfrom Uvo and Graham, (1998).
Table I. Summary of the results obtained from The Equatorial Pacific models for all sites available. COR Are the correlations between the series of normalized observed and estimated discharges obtained from the hindcast and the cross-validation of the Artificial Neural Network. All correlations are significant at 95% or more. SSE is the sum squared error for the whole series and SSEp, the SSE only for the years with anomalous normalized discharge ±I standard deviation, divided by the number of years used in this summation. Tropical Atlantic SST
Equatorial Pacific SST MAMJJA
MAMJJA
SONDJF
SON
DJF
SON
MAMJJA
MAMJJA
SON DJF
SON
DJF
SON
SITE
COR
SSE
COR
SSE
COR
SSE
COR
SSE
COR
SSE
COR
SSE
COR
SSE
COR
SSE
Balbfna
0 ...
541
0.91
7.93
0.96
3.25
0.88
10.13
0.86
12.50
088
13.51
0,92
7.23
0.84
14.08
Belo Monte
090
1011
088
10.46
0.93
8.80
082
15.67
0.80
9.73
091
9.25
0.87
11.02
096
347
Boa Esperanca
0....
5.22
0.95
4.45
522
0.79
17.39
0.92
7.50
0.93
6.74
0.97
2.86
0.95
485
Coaraq Nunes
098
524
0.95
5.11
0'" 0.96
4.31
0.89
9.49
0....
8.45
0.86
17.02
0....
576
0....
826
0.91
762
0.96
3.23
081
16.25
0.80
8.71
0.92
13.67
0'" 088
5.00
curue-una
11.04
0.97
2.54
Manaus
0.95
4.42
0....
5.00
0.93
6.83
0.89
9.72
0.95
4.92
0,91
10.02
0.96
392
0.90
9.54
Porteira
0.95
4.92
0....
6.61
0.92
7.35
0....
4.96
0.90
9.71
0.91
11.83
0.96
3.30
0.89
9.74
Samuel
0.90
8.73
0.67
11.62
089
10.25
0.87
11.51
0.95
4.86
0.91
11.88
0.80
988
0.92
6.77
Serra Ouebrada
094
579
0.94
5.30
0.87
11.76
0.87
11.24
0.94
588
0.89
9.33
089
929
0.85
12.71
Tucurul
0.94
5.59
0.92
7.18
0.89
10.80
0.78
18.40
0.95
5.31
0.78
18,81
0.95
4.15
0.87
11.98
301
Results
Training and validation The ANN described in the methodology was trained by differenttrainingsets composed by a series of SST and one of discharge.The SST seriestestedconsisted of differentseasonsor groupsof seasons for the SST anomaliesas describedbefore.The dischargedata and the ANN training resultsare presented for each station. The results are swnmarized in Table 1. For each of the four SST inputs it is given a) correlation coefficients betweenthe observednormalized dischargesand the normalizeddischargesobtainedfrom cross-validation of the trainedANN (COR) and b) SSE for the estimatedseries.All values ofSSE and COR in the Table may be comparedas the numberof time stepsin all the seriesis the same and both the SST anomaliesand the dischargesare normalized. Results show that ANN is a convenientmethod for forecasting discharge at all studied sites, but the accuracy of the trained ANN depends on the ocean and the SST seasons used in the training set. The highest correlationsbetweenobservedand estimatedseries (~0.95 and statistically significantat 99"10) and lowest SSEs ( 5.5) were found when usingeitherthe Pacificor the AtlanticOcean SST separately. When the MAM-JJA-SON-D.JF period for the Pacific Ocean SST was used in the training set, the correlation betweenobservedand estimateddischarge for CoaracyNunes was of 0.96 and the SSE was 5.24 and for Manaus and Porteiracorrelation of 0.95 and a SSE of 4.42 and 4.92, respectively. For the MAM-JJA-SON period,CoaracyNunesand Boa Esperanca both had correlation of 0.95 and a SSE of 4.45 and 5.11,respectively. For DJF,the cross-validation forCoaracy Nunes(Figure2a),Balbina(Figure 2b) and Curua-Unaresultedin correlation of 0.96 eachand SSE of 4.31,3.25 and 3.23,respectively. For the last SST period tested(SON), Porteira,despitea correlationof 0.94, had a SSE of 4.96, which also indicated thatthe validation wasaccurate. Coaracy Nuneshad correlation above0.95 for most of the SST periods testedin the PacificOcean and consistentlyhad the bestcorrelations among all tested stations, even thoughits SSE were not the lowestwhencomparedto otherstations withsimilarcorrelation for the same SST period. When AtlanticOcean SST was used in the trainingset, the cross-validation using the MAM-JJA-SONDJF SST period resulted in correlation of 0.95 betweenobserved and estimated dischargeat Manaus, Samueland Tucurui with SSE of 4.92,4.89 and 5.31,respectively. When the DJF period was used, Boa Esperancahad correlationof 0.97 and SSE of 2.66 (Figure 3a). Manaus and Porteirahad a correlation of 0.96 each withSSE of3.92 and 3.30,respectively, and Tucurui had 0.95 correlation and SSE of 4.15. For SON, Curua-Unahad correlation of 0.97 and SSE 2.54 (Figure3b), for Belo Monte the correlation was 0.96 and SSE 3.47 and BoaEsperanca had correlation of 0.95 and SSE 2.54.No correlationsequal to or greaterthan 0.95 were found when MAM-JJA-SONAtlanticSST period was part of the training set.
302
4
.,
El 2
4.3
0.96
.l!u
en
is 0
~ -2
z
-4
a) 1950
b) 1960
1970
1980
1990
1950
1960
Years
1970
1980
1990
Years
Figure 2.Timeseriesofnonnalized observed (dots) andestimated discharge (solid line) from thecrossvalidation usingPacific Ocean SSTanomalies fora) Coaracy Nunesandb) Balbina, using DJF SSTseason. Thecorrelation coefficient between thetwotimeseriesis shownat thetop rightof eachfigure andareallstatistically significant at >99%. The SSEisshown at thetop leftof each figure.
.,
4r-------------,
4r-~-----------,
~ 2
2.54
'§
is 0'· ~ -2 Z
is
~ -2 Z
c)
d)
-4'---~--------~---l
1950
0.96
'" 4.31 ~ 2 £
1960
1970
1980
-4L-...~_~_~
1990
1950
1960
Years
__
1970
~_~---.J
1980
1990
Years
Figure 3.Timeseries ofnonnalized observed (dots) andestimated discharge (solid line) from thecrossvalidation usingAtlantic Ocean SSTanomalies fora)Boa Esperanea at DJFseason and for b) Curua-Una at SONseason. Thecorrelation coefficient between thetwotimeseries is shownat thetop rightof eachfigure and isstatistically significant at >99%. The SSEisshownat thetop leftof eachfigure. 15 b)
10
10
20 30 Expenment
40
so
0
-4
-2
0
4
15 d)
10
0
-4
-2
4
Figure 4.a) Series of 50-estimation andb)histogram of thedifference observed value - estimated values for Balbina discharge in 1992 estimated from Pacific Ocean SST anomalies. c) seriesof 50estimation and d) histogram of the difference observed value - estimated values for Coaracy Nunesdischarge in 1992estimated fromthePacific Ocean SSTanomalies during DJF.Fora) andc) thedashed lineistilenormalized observed discharge, solidstraight lineis theaverage of theestimated values (solid broken line) anddotted lineisa zeroline.
303
The resultsobtainedshowed thatthe discharge at Boa Esperanea, which is locatedin the PamaibaBasin - NortheastBrazil,has high responseto both the Atlanticand PacificOcean.These resultscorroborate those obtainedby Uvo et al. (1997)who foundhighdependence on the precipitation at NortheastBrazil from both oceans SST. WithinAmazonia,generally the stationslocalized in northernsub-basins hadhigh correlations and low SSE whenthe PacificOcean SST anomalieswere partof the trainingset.Converselywhen the Atlantic Ocean SST anomalies were used, the stations located in the southernsub-basinshad high correlations and low SSEs. For all stations,the best results were obtained when using only one SST season as part of the trainingset.
Forecasting
The dependenceof the designedANN to the starting positionson the error hyperplane (startingweights) is a criticalconsideration whenusingthe ANN for realtime forecasting. We testeddifferentforecasting possibilitiesin order to better understand the potcntialsof the ANN. The ANN was used in a verification processto forecastdischarge. The ANN was trainedwith the same trainingset startingfrom 50 differentrandompositionson theerror hyperplane (weights). In each case, the discharge of a year not included in the training set was forecasted. Our main concern was to observehow the estimationof dischargevaries with the startingweights. Different results were obtained depending on the pair of sets used as input. Two examples of the verification are shown in Figure4. The figureshowsa seriesof50 verifications of dischargeat Balbina in 1992 estimated from DJF PacificOcean SST anomalies.The cross-validationfor the ANN trained with the same trainingset is shown in Figure2b. Figure4c showsthe 50 estimations for Coaracy Nunes discharge in 1992,also from DJF PacificOcean SST anomalies. The cross-validation for this case is shown in Figure 2a. From the firstexample it can be seenthat the verified discharge does not vary greatly(solidbroken line) and most of its values are belowzero (dottedline)as it is the observeddischarge(dashedline).The solid straightline showsthe averageof the 50 verified valuesand is not vCI)' far fromthe observedvalues. The error distributionrelated to these test series (Figure 4b) shows that the expectederror is approximately zero. The secondexampledisplaysa situationwherethe estimatedvalues(solid line)had the correctsign but oscillatedaround a value(solid straightline)considerably lower than the observed value(dashed line). The error distributionfor this case (Figure4d) shows that the expectationis about 1.5.
304
Conclusions We used ANN to validate and forecast discharge from ocean SST sites in the northeastern South America. The ANN was a suitabletechnique for the estimationof discharge at the studiedsites. The accuracyof the model dependedon the site, on theocean consideredand on the SST period used in the training set. Within the Amazon Basin, the discharge at the sites situated in northern sub-basins was better forecasted when the Pacific Ocean SST anomalies were used as input. On the other hand, sites locatedin southernsub-basinswerebetterforecasted when the input was AtlanticOcean SST anomalies. The validationof the trained nets gave correlationsbetween observed and estimated series that ranged from 0.77 in the worst case to 0.97 in the best case. The SSE for estimated series from validation were, in general, small. However, the dependence of tile ANN model on starting weights prevents the ANN from providing a preciseforecast, buteven so, it provedto be useful as an indicator for the seasonal discharge anomaly one or two seasons in advance. More development and further experiments with differentkinds of ANNs could, in the future, result in an ANN design and training technique less dependent on the starting weights.
References Brubaker, K. L., D. Entekhabi, and P. S. Eagleson, (1993). Estimation of continental precipitation recycling, 1. Climate, 6, 1077-1089. Brubaker, K. L., D. Entekhabi, and P. S. Eagleson, (1994). Atmospheric water vapor transport and continental hydrology over the Americas, J. Hydro!., 155,407-428. da Silva, A. M., C. C. Young and S. Levitus, (1994). Atlas of Surface Marine Data 1994,Volume 2: Anomaliesof DirectlyObservedQuantities. NOM AtlasNESDIS7, U.S. Departmentof Commerce, NOAA, NESDIS. Demuth, H. B. and M. Beale,(1994). Neural Network Toolbox for use with Matlah, The Math Works Inc. Derks E. P. P. A., M. L. M. Beckers,W. 1. Meissen and L. M. C. Buydens, (1996). Parallelprocesses ofchemical information in a local area network- H. A parallel cross-validation procedurefor artificial neural network, Comp. Chem. 20, 439-448. Dickinson, R. E. (Ed.), (1987). The geophysioJogy of Amazonia, vegetation and climate interactions, John Wiley and Sons, New York,pp. 1-526. Eagleson, P. S., (1986). The emergenceof global-scale hydrology,Water Resour. Res.,22, 6S-14S. French, M. N., W. F, Krajewski and R. R. Cuykendalf,(1992). Rainfall forecastingin space and time using a neural network, 1. Hydrol., 137, 1-31. Hagan, M. T., H. B. Demuth and M. Beale, (1995). Neural Network Design, PWS Publ. Company, Boston. Hastenrath, S. and L. Greischar, Circulation mechanisms related to Northeast Brazil rainfall anomalies, 1. Geophys. Res., 98(03), 5093-5102,1993. 305
HiII,T., M. O'Connorand W. Remus,(1996).Neural networkmodelsfor time series forecasts, Manag. Sci.,42, 1082-1092. Hsu,K.,H. V. Guptaand S. Sorooshian, (1995). Artificial neural networkmodeling of the rainfall-runoff process, Water Resour. Res., 31,2517-2530. Junk,W.1.and K. Furch,(1985). The physical settingand chemical properties of Amazonian watersand their relationships with the biota,In: Prance,G. T. and T. E. Lovejoy,(Eds.),Amazonia,Pergamon Press,pp. 3-15. Marengo,1. A., (1992).Interannual variability of surface climatein the AmazonBasin,Int.1.Clim., 12, 853-863. Marengo,1. A., L. M. Druyan and S. Hastenrath, (1993). Observational and modeling studies of Amazonia interannual climate variability, Clim. Change, 23, 267-286. Molion, L. C. B. (1975). A c1imatonomic study of the energyand moisturefluxesof Amazonas Basin with considerationof deflorestation effects.PhD Thesis,University of Wisconsing, Madison-W1. Molion, L. C. B. and 1. C; Moraes,(1987). RevistaBrasileira de Engenharia, Caderno de Hidrologia, 5,53-63. Mourn, A. D. and J. Shukla, (1981).On the dynamicsof droughts in Northeast Brazil:Observations, theoryand numericalexperiments with a generalcirculation model.1. Atmos. Sci., 38, 2653-2675. PoffN. L,. S. Tokar and P. Johnson,(1996).Stream hydrological and ecological responsesto climate change assessedwith an artificial neuralnetwork,LirnnoI.Oceanogr., 41, 857-863. Rao, V.B.and K. Hada,(1990).Characteristics of rainfall overBrazil: annualvariations and connections with the El Nifio-Southern Oscillation. Thcor.Appl. Climatol., 42, 81-91. Ropelewski, C. and M. Halpert, (1987). Globaland regional scaleprecipitation patternsassociated with the El Niiio/Southern Oscillation. Mon. Wea.Rev., 115, 1606-1626. Ropelewski, C. and M. Halpert,(1989). Precipitation patternsassociated with the high index phase of the SouthernOscillation. 1. Climate,2, 268-284. Salati,E. and J. Marques,(1984).Climatology of the Amazon regionin The Amazon Limnology and LandscapeEcologyof a MightyTropicalRiverand its Basinedited by H. Sioli,The Hague Boston W.Junk. Salati,E., (1986).The climatology and Hydrology of Amazonia. In: Amazonia. G. T. Pranceand T. M. Lovejoy(Eds.): 18-48. PcrgamonPress,Oxford,442 pp. Uvo, C. B., C. A. Repelli,S. E. Zcbiakand Y. Kushnir,(1997).The influenceof tropical Pacific and AtlanticSST on NortheastBrazilmonthlyprecipitation. In press1. Clim. Uvo, C. B. and N. E. Graham, (1998). Seasonal runoffforecast forNorthernSouthAmerica: a statistical model,(subm.to Water Resour. Res.,)
306
Program.
Appendix 1
Sunday 28 June, Arrival, small get-together at Hotel Concordia 18:00-19:00. Monday 29 June; - Gathering at Hotel Concordia for walking to Grand Hotel. 09:0009:20-09:30 - Introduction (Dr. Ronny Bemdtsson, LU, and Dr. Jean Albergel, ORSTOM).
Observation techniques; GIS/remote sensing; climatic, soil, agronomic, and socioeconomic data storage and processing for small watersheds; (chairman: Dr. Jean Albergel, ORSTOM). 09:30-10:00 10:00-10:30 10:30-11 :00
I1 :00-11 :30 11:30-12:00
12:00-12:30 12:30-13:00 13:00-14:00
- New developments in the Wadi Hydrology Network, Dr. Jean Khouri, ACSAD. - The Sindyaneh Wadi Basin in Syria, Dr. Abdallah Droubi, Mr. Yasser Ibrahim, ACSAD and Or. Jean Albergel, ORSTOM. - Small dams' water balance: experimental conditions, data processing and modeling, Dr. Jean Albergel, ORSTOM, Mr. Slah Nasri, INRGREF, and Dr. Mohamed Boufaroua, MAT. - Coffee and tea break. -Tntegrating soil profile and soil hydraulic properties data bases to be used in simulation models and land evaluation expert systems, Prof. Felix Moreno, Dr. D. de la Rosa, and Dr. J. E. Fernandez, IRNASE. - Lebanese hydrology and needs for water storage, Dr. Bassam Jaber and Dr. Fuad Saad, MHER. - Remote sensing applications for the management of small catchments in arid and semiarid area, Dr. Chuqun Chen, CAS. - Lunch
Water quality and quantity: hydrological and transport modeling; (chairman: Dr. Jean Khouri, ACSAD). 14:00-14:30 14:30-15:00
15:00-15:30 15:30-16:00 16:00-16:30 16:30-17:00 19:0021:00-
- Water chemistry characteristics in small reservoirs of the semiarid Tunisia, Or. Nathalie Rahaingomanana, ORSTOM. - Water chemistry of small reservoir catchments in central Tunisia, Dr. O. Grunberger, Dr. Jean-Pierre Montoroi, ORSTOM, Mr. Slah Nasri, INRGREF, Or. Jean Albergel, Mr. Yannick Pepin, and Dr. Nathalie Rahaingomanana, ORSTOM. - Modelling effects of preferential flow and transport on non-point source pollution, Prof. Nicholas Jarvis, SLU. - Coffee and tea break. - Solute transport and soil water content measurements in arid soils using time domain reflectometry, Dr. Magnus Persson, LU. - The FLAPS system for data management and hydrological modeling, Dr. Linus Zhang, LU. - Dinner - Optional software demonstration (Hotel Concordia). 307
Tuesday 30 June; Rainwater harvesting; infiltration techniques and modeling: infiltration and erosion (chairman: Dr. Nejib Rejeb, INRGREF). 09:00-09:30 - Microcatchmcnt, macrocatchment, and flood water harvesting techniques applied in the Mediterranean, Prof. Dr. Dieter Prinz, KU. 09:30-10:00 - The use of TOR for wetness measurements in soil erosion and conservation practices in small watersheds, Or. Patrick Zante, ORSTOM, and Mr. Slah Nasri, INRGREF. 10:00-10:30 - Coffee and tea break. 10:30-11 :00 - Land use transformation impact on reservoir siltation in Morocco: the need for better assessment tools, Dr. Abdelaziz Merzouk, lA V. I1:00-11 :30 - Modeling small dams' siltation with MUSLE, Or. Jean Albergel and Mr. Yannick Pepin, ORSTOM. 11:30-12:00 - Small-scale cistern system for rainwater collection and storage in northwestern China, Or. Linus Zhang, LU and Prof. Kun Zhu, LRl, 12:00-12:30 - Disinfection and fresh-keeping of rainwater in small scale cisterns, Prof. Kun Zhu and Or. Chen Hui, LRI, Or. Linus Zhang and Or. Ronny Berndtsson, LU. 12:30-13:00 - Strategy of soil and water conservation in Tunisia, Or. Habib Farhat and Dr. Mohamed Boufaroua, MAT. 13:00-13:30 - Lunch 14:00-16:00 - Guided round trip in Lund on foot, meeting place: Cathedral entrance. 18:00- Optional software demonstration (Hotel Concordia) Wednesday I July; Reservoir planning, operation and management; Rainfall-inflow relationships; Dam design and operation; Surface-groundwater interactions (chairman: Dr. Abdelaziz Mereouk.TAv). 09:30-10:00 - Hydrodynamics and related water quality in small reservoirs, Prof. Lars Bengtsson, LU. 10:00-10:30 - Coffee and tea break. 10:30-11 :00- Groundwater recharge and modeling in an experimental catchment, Mr. Slah Nasri, lNRGREF. 11:00-11 :30 - Deterministic versus stochastic hydrological modeIing; uncertainties and decisions, Mr. Jan Hoybye, LU. 11:30-12:00 - Neural network methodology to simulate discharge, Or. Cintia Uvo, LU. 12:00-12:30 - Water reservoir management in practice, Dr. Laszlo Iritz, LU. 12:30-13:30 - Lunch 14:00-17:00 - Study visit to the Vomb water works and the Scania region. 18:00- Optional software demonstration (Hotel Concordia)
308
Thursday 2 July; Study tour to Denmark and Copenhagen. 08:00- Departure Hotel Concordia 09:00- Short stop in Helsingborg and the Karnan castle 09:30- Ferry to Denmark. - The Cronborg Castle (the Hamlet and Shakespeare story) 10:1512:45- Danish typical lunch 14:30- Free time in Copenhagen 16:00 - Departure for Malmoe and Sweden
Abbreviations: ACSAD: CAS: IAV: INRGREF: IRNAS: KU: LRI: LU: MAT: MHER: MIDS: ORSTOM: SLU:
Arab Center for the Studies of Arid Zones and Dry Lands, Syria. South China Sea Institute of Oceanography, Chinese Academy of Sciences, China Institute for Agronomy and Veterinary Hassan H, Morocco. National Institute for Research on Rural Engineering, Water, and Forestry, Tunisia. Institute for Natural Resources and Agrobiology, Spain. Karlsruhe University, Germany. Lanzhou Railway Institute, China. Lund University, Sweden. Ministry of Agriculture, Tunisia. Ministry of Hydraulic and Electric Resources, Beirut, Lebanon. Ministry of Irrigation, Damascus, Syria. French Institute for Scientific Research and Cooperative Development, France/Tunisia. Swedish Agricullural University, Uppsala, Sweden.
309
310
Participant List:
Appendix 2
Dr. Jean AlbergeJ Programme HYDROMED Mission ORSTOM B.P. 434 1004 Tunis, El Menzah, Tunisia Fax: +216-1-750 254 Email:
[email protected] Prof. Lars Bengtsson Dep. of Water Resources Engineering Lund University, Box 118 S-221 00 Lund, Sweden Fax: +46-46-22244 35 Email:
[email protected] Dr. Ronny Berndtsson Dep. of Water Resources Engineering Lund University, Box 118 S-221 00 Lund, Sweden Fax: +46-46-22244 35 Email:
[email protected] Dr. Mohamed Boufaroua Soil Conservation Directorate Ministry of Agriculture Tunis, Tunisia Dr. Chuqun Chen Remote Sensing Group South China Sea Institute of Oceanography Chinese Academy of Sciences 164 West Xingang road Guangzhou 51030 J, China Fax: +86-20-84451672 Email:
[email protected] Dr. Salah Qara Damour Ministry of Irrigation Damascus, Syria Fax: +963-11-532 3063 Email:
[email protected]
311
Dr. Abdallah Droubi ACSAD, Arab Center for the Studies of Arid Zones and Dry Lands Division des Ressourees en Eau B.P.2440 Damascus, Syria Fax: +963-11-532 3063 Email:
[email protected] Dr. Habib Fahrat Soil Conservation Directorate Ministry of Agriculture Tunis, Tunisia Mr. lan Hoybye Dep. of Water Resources Engineering Lund University, Box 118 S-221 00 Lund, Sweden Fax: +46-46-2224435 Email:
[email protected] Mr. Yasser Ibrahim ACSAD, Arab Ccnter for the Studies of Arid Zones and Dry Lands Division des Ressources en Eau B.P.2440 Damascus, Syria Fax: +963-11-532 3063 Email:
[email protected] Dr. Laszlo Iritz Dep. of Water Resources Engineering Lund University, Box 118 S-221 00 Lurid, Sweden Fax: +46-46-222 4435 Email:
[email protected] Dr. Bassam laber Ministry of Water and Electric Resources Autostrade El Nahr Beirut, Lebanon Fax: +961-1-565555
312
Prof. NichoJas Jarvis Dep. of Soil Science Swedish Agricultural University Box 7014 S-750 07 Uppsala, Sweden Fax: +46-18-67 27 95 Email:
[email protected] Dr. JeanKhouri ACSAD, Arab Center for the Studies of Arid Zones and Dry Lands Division des Ressources en Eau B.P.2440 Damascus, Syria Fax: +963-11-532 3063 Email:
[email protected] Or. Mohamed Mejjati lAV, Lab. des Sciences du Sol B.P. 6202-instituts Rue Allal Fassi, Al Irfane Rabat, Maroc Fax: +212-7-771285 Email:
[email protected] Or. Abdelaziz Merzouk lAV, Lab. des Sciences du Sol B.P. 6202-instituts Rue AlIal Fassi, AI Irfane Rabat, Maroc Fax: +212-7-771285 Email:
[email protected] Dr. Jean-Pierre Montoroi Centre ORSTOM d'lle-de-France Laboratoire des Formations Superficiellcs 32 avenue Henri Varagnat 93143 BONDY Cedex, France Fax: +33 - 1 - 48 47 30 88 email:
[email protected]
313
Prof. Felix Moreno IRNAS, Institute for Natural Resources and Agrobiology r.o. Box 1052 Sevilla 41080, Spain Fax: +34-5-462 4002 Email:
[email protected] Mr. Slah Nasri INRGREF Route de la Soukra B. P. No. 10 Ariana Tunis, Tunisia Fax: +216-1-717951 Email:
[email protected] Dr. Magnus Persson Dep, of Water Resources Engineering Lund University, Box 118 S-221 00 Lund, Sweden Fax: +46-46-222 4435 Email:
[email protected] Prof. Dr. Dieter Prinz Dep. of Rural Engineering University of Karlsruhe 0-76128 Karlsruhe, Germany Fax: +49-721-608-6165 (661634) Email:
[email protected] Dr. Nathalie Rahaingomanana ORSTOM Laboratoire du Garnet Avenue Agropolis 34045 Montpellier, France Email:
[email protected] Or. Nejib Rejeb INGREF Route de la Soukra B. P. No. 10 Ariana Tunis, Tunisia Fax: +216-1-717951
314
Or. Fuad Saad Ministry of Water and Electric Resources Autostrade El Nahr Beirut, Lebanon Fax: +961-1-565555 Ms. Pernilla Somogyi Dep, of Water Resources Engineering Lund University, Box 118 S-221 00 Lund, Sweden Fax: +46-46-222 4435 Email:
[email protected] Or. Cintia Uvo Dep, of Water Resources Engineering Lund University, Box 118 s-n I 00 Lund, Sweden Fax: +46-46-222 4435 Email:
[email protected] Or. Patrick Zante Programme HYOROMEO Mission ORSTOM B.P. 434 1004 Tunis, El Menzah, Tunisia Fax: +216-1-750254 Email:
[email protected] Or. Linus Zhang Dep. of Water Resources Engineering Lund University, Box 118 S-221 00 Lund, Sweden Fax: +46-46-222 44 35 Email: Linus.Zhangrmtvrl.lth.se Prof. Kun Zhu Water Resources and Environmental Technology Research Center Oep. of Environmental Engineering Lanzhou Railway Institute, China Fax: +86-931-776 76 61 Email:
[email protected]
315
Tryckt
pa Vag & Vatten 1999
